(navigation image)
Home American Libraries | Canadian Libraries | Universal Library | Community Texts | Project Gutenberg | Children's Library | Biodiversity Heritage Library | Additional Collections
Search: Advanced Search
Anonymous User (login or join us)
Upload
See other formats

Full text of "High Vacuum Technology"

£1) 

-? 
Q. 

C 

ro 

CD 

c 

Z3 
Z3 



p introduction to the theory and practice of 

HIGH VACUUM TECHNOLOGY 



G) 

o 

c 
c 



o 

Q. 

c 
o 

I— I- 

o' 







o g 



O 



Q. 






o 





621. 

55 
WAR 



L.Ward J.PBunn 



NOT FOR 
LOAN ^ 



4^ 




INTRODUCTION TO 

THE THEORY AND PRACTICE OF 

HIGH VACUUM TECHNOLOGY 



L. WARD, M.Sc, Ph.D., F. Inst. P. 

Principal Lecturer in Applied Physics 
Lanchester College of Technologp, Coventry 

and 

J. P. BUNN, M.Sc, A. Inst. P. 

Senior Lecturer in Applied Physics 
Lanchester College of Technology, Coventry 



u 



LONDON 
BUTTERWORTHS 

1967 



ENGLAND: 



AUSTRALIA: 



CANADA: 



NEW ZEALAND: 



SOUTH AFRICA; 



U.S.A.: 



BUTTERWORTH & CO. (PUBLISHERS) LTD. 
LONDON: 88 Kingsway, W.C.2 

BUTTERWORTH & CO. (AUSTRALIA) LTD. 
SYDNEY: 20 Loftus Street 
MELBOURNE: 473 Bourke Street 
BRISBANE: 240 Queen Street 

BUTTERWORTH & CO. (CANADA) LTD. 
TORONTO: 1367 Danforth Avenue, 6 

BUTTERWORTH & CO. (NEW ZEALAND) LTD. 
WELLINGTON: 49/51 Ballance Street 
AUCKLAND: 35 High Street 

BUTTERWORTH & CO. (SOUTH AFRICA) LTD. 
DURBAN: 33/35 Beach Grove 

BUTTERWORTH INC. 

WASHINGTON, D.C.: 7300 Pearl Street, 20014 









te 



ty 



V>fe"7&i-<^^^ 



Butterworth & Co. (Publishers) Ltd. 
1967 



M:%1' Suggested U.D.C. number: 533-5 



Suggested additional numbers: 621 — 186-2 

621-52 



Made and Printed in Great Britain by William Clowes and Sons, Limited 
London and Beccles 



PREFACE 



In preparing this book we have attempted to present a general 
survey of the problems, the basic science and the technology involved 
in the production of vacua. The emphasis has been placed on the 
underlying physical principles rather than on detailed discussions of 
specialized items of equipment. We have tried to indicate the general 
lines upon which vacuum technology seems to be developing but it 
has not been our intention to provide a comprehensive review of the 
latest research. 

The first two chapters are devoted to fundamental properties of 
gases and vapours and their flow behaviour. Then follow chapters 
on the basic components of vacuum systems, including piunps, 
gauges, couplings and valves. The use of basic concepts and of the 
components is then illustrated in chapters dealing with the design, 
construction, operation and leak testing of practical systems. In the 
final chapter brief surveys are given of some of the fields of application. 
These are not intended to be exhaustive but to demonstrate further 
the way in which fundamental ideas are applied in practice. The 
subject of ultra-high vacuum has not been presented as a separate 
topic but as a logical development of high vacuimi technique since, 
although there are many practical refinements, the same basic 
principles are applicable. 

It is hoped that this book will provide a suitable background for 
both graduates and undergraduates in Universities and Technical 
Colleges, and also an adequate introduction for industrial and 
laboratory workers faced with their first experience of high vacuum. 
In addition the more experienced worker should gain a greater under- 
standing of the techniques and equipment which he uses. 

It would be impossible to acknowledge all those who have given us 
help in preparing this book but, in addition to those specifically 
acknowledged in the text, we would express our gratitude to MuUard 
Ltd., British Lighting Industries, and Imperial Metal Industries Ltd. 
for their assistance in preparing various sections of Chapter 8; 
to Edwards High Vacuum Ltd., Birvac Ltd., Genevac Ltd. and 



PREFACE 

Leybold's Nachfolger from whose technical literature we have obtained 
data and the bases of many diagrams; to our publishers who have 
given us patient and understanding assistance. 

Finally, we extend our apologies to any whose work we have quoted 
or used and inadvertently failed to acknowledge. 

L.W. 
J.P.B. 



VI 



CONTENTS 

Preface .......... v 

1 Basic Vacuum Concepts ...... 1 

2 Theory of Gaseous Flow ...... 24 

3 The Production of Vacuum . . . . .37 

4 Measurement of Low Pressure . . . . .71 

5 The Construction of Vacuum Systems . . . .111 

6 The Operation and Design of Vacuum Systems . .134 

7 Leaks AND Leak Detection . . . . .158 

8 Some Applications of Vacuum Technique . . . 1 70 
List of Principal Symbols ...... 207 

Index 211 



Vll 



BASIC VACUUM CONCEPTS 



1.1 Introduction 

The aim of vacuum technologists is to remove from an enclosed space 
as much gas or vapour as is necessary for a particular purpose. In the 
majority of cases this gas is air, but occasionally other gases are 
pumped; also gases and vapours are very often evolved within the 
enclosed space. Whatever its origin, however, the material removed 
is always in the gaseous state and this chapter is a brief discussion of 
the concepts involved in the study of gases. 

1 .2 Pressure Units 

The unit of pressure used in vacuum work is the torr, which is de- 
fined'i' as one-760th part of standard atmospheric pressure {i.e. of 
1,013,250-00 dyn cm~^). However, pressure is frequently measured 
in terms of the height [l mm) of the column of liquid which it can 
support. If the liquid density is p g cm~^ then the absolute pressure 
is given by pglj 10 dyn cm"^, where g is the standard acceleration due 
to gravity (980-665 cm sec"^)'^^ Thus the absolute pressure due to a 
colimin of mercury (with p = 13-5951 gcm"^)<^^ 760 mm high is 
1,013,250-14 dyn cm"^, and it can be seen that 1 torr is numerically 
almost the same as a pressure of 1 mm of mercury. 



1.3 Gas Laws 

The fundamental relationship between the pressure /> of a gas and the 
volume V it occupies is expressed by the ideal gas equation 



pV = NkT 



(1.1) 



where N is the total number of molecules present, T is the absolute 
temperature, and k is Boltzmann's constant (the gas constant per 
molecule, numerically equal to 1-38 x 10" ^^ erg "K"^ or 1-03 x 10"^^ 
torr 1. °K ~ ^) . If the quantity of gas present is one gramme molecule 



1 I.H.V.T. 



I 



BASIC VACUUM CONCEPTS 

( 1 mole) then N = Na (Avogadro's number, numerically equal to 
6-02 X 10^^) and kNa = R, the gas constant per mole, which has the 
value 2 calmole-^ °K-i or 8-31 x IC erg mole "^ °K-i. 

At constant temperature and for a fixed mass of gas, the product 
pVis constant, which is a mathematical expression of the well-known 
Boyle's Law. 

1 .4 Throughput and Speed 

In vacuum systems it is necessary to define a measure of the rate of 
flow of gas. Ideally this should be in terms of the mass flowing per 
unit time, but in practice it is not convenient to measure gaseous 
masses directly. However, eqn 1 . 1 shows that at constant temperature 
the mass of gas in a given system is proportional to the product of 
pressure and volume (both of which are fairly easily measured) . The 
mass rate of flow is proportional to the time differential oipV, and this 
fact is used in vacuum practice to measure gas flow rate g which is 
then defined as 



dt 



ipv) 



(1.2) 



This quantity is called throughput. By using eq 1.1, an alternative 
expression for the throughput may be shown to be 






(1.3) 



The negative sign is used to give positive values of g, since (d/di)(/)F) 
and dNIdt are usually negative. 
Eqn 1 .2 may also be expressed as 



dt ^ dt 



(1.4) 



If Fis constant, g = — Vdpldt; this situation holds in a vessel of fixed 
volimie and the throughput is measured in terms of the rate of fall of 
pressure. 

At a point in the system where the instantaneous pressure is 
p, g = —p dVldt; the throughput is then measured in terms of the 
rate of flow of gas volume (measured at pressure p) past the point. 
The term —dVjdt is called the speed S, at the point and hence 

g = Sp (1.5) 

The concept of speed is particularly useful when dealing with vacuum 
pumps, where the speed of the pump is defined as the volume of gas 

2 



KINETIC THEORY 



removed from the system per unit time, measured at the pressure at 
the pump inlet. 

With the units used throughout this book, throughput is measured in 
torr 1. sec"^ and speed in 1. sec~^ or 1. min"^. 



1.5 Kinetic Theory 

The gaseous state is the easiest of the three states of matter to deal with 
theoretically. The basic theory has been developed for almost a 
century and is known as the Kinetic Theory. 

All three states of matter exist as assemblies of molecules which, for 
the purposes of this discussion, may be regarded in the classical form 
as simple elastic spheres. It is also assumed that (f) the molecules 
move freely and randomly throughout the volume of the containing 
vessels, (if) the actual volume occupied by the spherical molecules is 
negligible compared with the volume of the space they occupy, and 
{in) the forces between the gas molecules are negligible. It is found 
that a gas which fulfils these conditions obeys the ideal gas eqn 1.1. 

These conditions apply at atmospheric pressure and obviously 
become no less valid as the pressure is reduced, although if the pressure 
is increased above atmospheric there will come a point when these 
simple ideas need modification. 

1.5.1 Gas Pressure 

Gas molecules make collisions both with the walls of the containing 
vessel and with other gas molecules. The pressure produced by a gas 
is the direct result of the molecular impacts on the containing walls. 
By using the Kinetic Theory (see for example Kennard'"') the 
pressure may be shown to be 

p = \mnc^ (1.6) 

where m is the mass of each molecule, n = NjV is the molecular 
density (number of molecules per unit volume) and c is the root mean 
square velocity of the molecules defined by 

where Wj is the number of molecules having velocity fj. 
Combining eqns 1.1 and 1.6 gives 



BASIC VACUUM CONCEPTS 



Hence 



. = (^) (1.8) 

It is also possible to define an arithmetic mean velocity c^ as 



" 2«. 



and it may be shown that 



SkTV"^ 



\ Tim J 



(1.9) 



1.5.2 Molecular Bombardments 

The number of molecular bombardments v on unit area of wall in 
unit time can be deduced from Kinetic Theory'*' as 

V = inca 



Using eqns 1 .6, 1 .8 and 1 .9 it is easily shown that 

= 3-50 X 1022j!)(Mr)-i'2mol./cm-2 sec- 1 



(l.IO) 

(1.11) 
(1.11a) 



where p is in torr and M is the molecular weight in grammes. 

The quantity of gas q coming from a region at pressure p and 
striking an area A in unit time, is equal to vAkT, i.e. 



^=^^(l^)"'"^°^-' 



•kTVi'' 



torr 1. sec ^ 



(1.12) 
(1.12a) 



if j& is in torr, A in cm^ and M the molecular weight in grammes. 



1.5.3 Mean Free Path 

Intermolecular collisions occur more frequently when the molecules 
are more densely packed. The average distance travelled by a gas 
molecule between successive collisions is called the mean free path A. 
Once again, there are standard derivations in Kinetic Theory for 



DALTON S LAW OF PARTIAL PRESSURES 

evaluating the mean free path, and here it is only necessary to quote 
the result*^' 



1 



(1.13) 



where a is the molecular diameter. 

As A is proportional to 1/n and pressure is proportional to n it 
follows that the mean free path is inversely proportional to pressure. 
The final column in Table 1.1 gives the approximate values of mean 
free path for different pressures of air. 

With ordinary laboratory equipment, when the pressure is reduced 
to about 10 ~* torr the mean free path is comparable with the dimen- 
sions of the apparatus. At lower pressures the gas molecules collide 
more frequently with the walls of the vessel than with one another 
which, as is seen later, leads to marked differences in gas behaviour 
compared with that at higher pressures. 

1.5.4 Summary 

The molecular density, the number of bombardments per unit area 
per second and the mean free path are given in Table 1.1 for air at 
20°C and various pressures. 



Table I.l. Kinetic Theory Data for Air at 20°C 



Pressure 
(torr) 


Molecular 

density 

(mol. cm-3) 


No. of bombardments 1 
unit areajsec 
(cm-''sec-i) 


Mean free path 
(cm) 


760 
1 
10-3 
10-8 
10-8 
10-1=" 


2-5x1018 

3-3x1018 

3-3x1013 

3-3x101° 

3-3x10'' 

3-3x10* 


3-2 X 10=3 
4-3x10=° 
4-3 X 101'' 
4-3x101* 
4-3x1011 
4-3x108 


6-5x10-8 
5 xlO-3 
5 X 10° 
5 X 10-^3 
5 xlO + 8 
5 xlO + 8 



1 .6 Dalton's Law of Partial Pressures 

When there is a mixture of gases, as in air, each component produces 
Its own pressure known as the partial pressure. The total pressure 
is then the sum of the partial pressures. This statement is known as 
Dalton's Law of Partial Pressures and is especially useful when dealing 
with mixtures of gases and vapours. 



BASIC VACUUM CONCEPTS 



1.7 Viscosity of Gases 



The viscosity of a gas may be regarded as a measure of its resistance to 
flow and the concept of viscosity is thus of great importance in 
vacuum work. The detailed study of the flow of gases is carried out in 
Chapter 2, but a general idea of gaseous viscosity may be obtained 
from the following argument. 

Consider a tube containing gas whose mean free path is very much 
less than the tube diameter, and let there be a pressure difference 
between the ends of the tube. There is a force acting on the gas from 
the region at high pressure to that at low pressure. The gas does not 
behave like a rigid body and different parts of the gas move with 
difTerent velocities. In particular, it is an experimental fact that the 
layer of gas molecules immediately adjacent to the wall of the tube 
does not have any component of velocity along the tube. One way of 
explaining this is to say that this layer of molecules experiences a 
frictional force equal and opposite to the force due to the pressure. 
The next layer experiences the pressure force and also a smaller force 
in the opposite direction due to friction with the first layer. The 
resultant force causes the second layer of gas molecules to slide for- 
ward along the tube; a similar argument can be applied to all other 
layers of gas. It is found that the gas velocity increases to a maximum 
value at the centre of the tube. Such behaviour is termed laminar 
flow with layers of gas at equal distances from the wall travelling with 
the same velocity. 

The value of the viscous drag, i.e. the frictional force per unit area, is 
found to be proportional to the velocity gradient perpendicular to the 
direction of motion. The constant of proportionality is termed the 
coefficient of viscosity and is denoted by rj. 

Viscosity may also be interpreted in terms of the Kinetic Theory. 
In addition to the velocity along the tube produced by the pressure 
difference, the individual gas molecules possess velocities which are 
distributed randomly in direction. There is consequentiy a con- 
tinuous molecular movement from one layer to another; molecules 
moving from the centre towards the walls travel to layers of smaller 
average momentum and vice versa. Intermolecular collisions ensure 
the redistribution of this momentum, and viscosity may be regarded in 
the Kinetic Theory as the transport of momentum. 

It can be shown*^' that tj = ^nmXc^ and substituting the expressions 
for n and A it follows that 



= 3^7^ 
6 



THERMAL CONDUCTIVITY 

Thus the viscosity is independent of pressure but proportional to \/ T. 
As the pressure is decreased the mean free path eventually becomes 
greater than the diameter of the tube and the simple picture of 
laminar flow is no longer tenable. A new type of flow called molecular 
flow is produced, in which the gas molecules strike the walls more 
frequently than they collide with one another. Under these condi- 
tions the concept of viscosity as defined above is no longer valid. 

1.8 Thermal Conductivity 

By treating it in a similar manner to viscosity, the thermal conductivity 
of a gas"^ can also be calculated. When a temperature gradient exists 
in a gas, molecules from the hot region, possessing high kinetic 
energies, travel to the colder part and dissipate their excess energy by 
intermolecular collisions. Using the Kinetic Theory it can be shown 
that thermal conductivity k = ^nmXC^, where C„ is the specific heat of 
the gas at constant volume. 

Again at high pressures, when A is small, k is independent of pres- 
sure. At low pressures where A is large and energy dissipation by 
intermolecular collisions is negligible it is found that k is proportional 
to pressure. This enables a measurement of thermal conductivity 
to be used in giving an estimate of the gas pressure, and is the principle 
of operation of the Pirani gauge discussed in Chapter 4. 

1.9 Vapours 

In addition to pumping gases, a vacuum system has also to deal with 
vapours. The distinction between a gas and a vapour is a subtle one 
and is considered in more detail later in this section. 

A vapour is produced by the evaporation of a liquid and the process 
may be described in terms of the Kinetic Theory. Some of the 
molecules near the liquid surface have sufficient kinetic energy to 
escape into the atmosphere and exist as a ' gas '. Raising the tempera- 
ture facilitates this process. If the liquid is in the open, the vapour 
molecules rapidly diffuse away from the parent liquid and in general 
produce what is known as an unsaturated vapour. On the other 
hand, if the liquid is in an enclosed vessel there is a high probability 
that a free molecule will collide with the liquid surface and be recap- 
tured {i.e. condensed). Thus, a dynamic equilibrium will be estab- 
lished between evaporation and condensation in which the net number 
of free molecules in the gaseous state is constant. Under these 
conditions, the vapour is said to be saturated and the pressure it 
exerts is called the saturation vapour pressure (s.v.p.). Since the rate 



1 



BASIC VACUUM CONCEPTS 

of evaporation falls with decrease in temperature, the s.v.p. also 
decreases. 

If an unsaturated vapour is compressed at constant temperature, 
its pressvire and volume change approximately in accordance with 
Boyle's Law until the s.v.p. is reached. Further attempts at compres- 
sion cause vapour to condense leaving the remaining vapour at its 
s.v.p., a situation which obtains until the whole of the vapour is 
condensed. Thus, under conditions of saturation, Boyle's Law does 
not hold. This behaviour, however, does not occur at all tempera- 
tures. If the compression process is carried out at successively higher 
temperatures, eventually a temperature is reached (known as the 
critical temperature) at and above which the vapour will not con- 
dense, no matter what pressure is applied, i.e. the free molecules behave 
as a gas. Thus, free molecules behave as gases or vapours according 
as to whether they are at a temperature above or below their critical 
temperature. The so-called permanent gases (e.g. Hj, Og, Ng and 
the inert gases) have critical temperatures well below normal room 
temperature. On the other hand, the vapours of water, mercury, 
and the common organic cleaning liquids {e.g. acetone, benzene, 
alcohol, carbon tetrachloride) have critical temperatures above normal 
room temperature and hence can be liquefied by compression. 

An understanding of the properties of vapours is of great im- 
portance to a vacuum technologist, for the following reasons: 

(a) Any liquid surface inside a vacuum system is a source of vapour 
and, as long as any liquid remains, the minimum pressure attainable 
is the s.v.p. of that liquid. The implication of this is that great care 
must be taken to avoid contaminating a system {e.g. by careless 
handling or unclean work) with possible vapour sources. 

{b) A reduction in temperature of any part of the system reduces the 
s.v.p. of any vapours present throughout the system. This is the 
principle underlying the use of cold traps, refrigerated baffles, 
cryogenic pumps, etc., which are discussed later. 

{c) Certain pumps and pressure gauges depend for their action 
upon high gas compression. Vapours which, in the body of the 
system, are unsaturated and of low pressure may be condensed under 
these high compressions leading to faulty action of the pumps and 
gauges. 

These effects are discussed in detail in the relevant chapters. 

1.10 Ionization 

In the preceding discussion on gaseous properties it was sufficient to 
treat the gas molecules as solid elastic spheres. However, in order to 



IONIZATION 

understand some of the processes involved in producing and measur- 
ing low pressures it is necessary to examine the structure of the 
molecules. 

Gas molecules consist of groups of atoms bound together by 
relatively strong forces so that the molecular entity stays intact 
throughout all the intermolecular collisions. The atoms consist of a 
positively charged nucleus round which circulate a number of 
electrons carrying a total negative charge equal in magnitude to that 
on the nucleus, so that the atom as a whole is electrically neutral. 
Almost all the mass of the atom is concentrated in the nucleus which 
is made up from two fundamental particles, protons and neutrons, 
each of about the same mass. The protons are positively charged but 
the neutrons are uncharged. The chemical properties of an atom 
depend on the nuclear charge but not on the nuclear mass and thus, 
by adding or removing neutrons from the nucleus, it is possible to 
have atoms of the same chemical element but of different masses. 
Such atoms are called isotopes of the element and are indicated 
symbolically by their chemical symbol with their mass number as a 
suffix in the superior position. For example, the isotope of nitrogen 
of mass number 14 is represented by N^*. 

If one or more of the circulating electrons of an atom or molecule 
is removed, the atom is left positively charged and is then known as a 
positive ion. In order to produce ionization, sufficient energy must 
be supplied to one of the electrons for it to overcome the electrostatic 
forces binding it to the nucleus. This quantity of energy is referred 
to as the ionization energy and varies from element to element. 
Single ionization {i.e. the removal of a single electron) is the most 
common process but it is also possible to produce multiple ionization 
The degree of ionization is indicated by the chemical formula of the 
molecule with the appropriate number of + signs as a suffix in the 
superior position. For example, doubly ionized carbon monoxide is 
represented by CO ■*" ■*■ . 

There are several ways of producing ionization, such as bombarding 
the atoms with (f ) free electrons of kinetic energy at least equal to the 
ionization energy, {ii) radiation from radioactive materials, or {Hi) 
cosmic radiation. Of these, the first is the most common method used 
in vacuum processes and instruments. 

Once a gas has been ionized the motion of the ions can be in- 
fluenced by electric and magnetic fields. 

Consider an ion of charge e and mass m,, in an electric field of 
strength E. The force on the ion due to the electric field is eE, and 
this causes acceleration of the ion along the direction of the field. 
The velocity of the ion continues to increase until the ion collides either 



BASIC VACUUM CONCEPTS 

with another atom or with the electrode producing the iield. In the 
case of collision with other atoms, if the kinetic energy of the ion on 
impact is much greater than the ionization energy of the atom, 
secondary ionization may take place. 

Alternatively, if the ion moves with a velocity v perpendicular to a 
magnetic field of flux density B, it experiences a force equal to Bev in a 
direction perpendicular to both B and v. This force causes the ion to 
travel in a circular path of radius r, where r = m^vlBe. 

Similar arguments can be applied to the motion of electrons in 
electric and magnetic fields. 

1.11 Sorption and Desorption 

In the discussion on the simple Kinetic Theory of gases, it was as- 
sumed that the only interactions between gas molecules and the walls 
of the containing vessel are simple elastic collisions. In fact other 
types of interaction occur which have a profound effect upon the 
degree of vacuum obtainable and also upon the processes used in 
achieving it. One group of these interactions can be classified under 
the general heading of sorption. It is convenient to consider sorption 
as two separate processes, namely adsorption and absorption, al- 
though in practice both processes usually occur together. 

1.11.1 Adsorption 

This refers to the process whereby molecules are attracted to and 
become attached to the surface of a solid, the resulting layer of 
adsorbed gas being a few molecules thick. The attracting forces of 
the solid may be physical or chemical. Since the forces are attrac- 
tive, work is done in adsorbing molecules and heat is generated; to 
desorb molecules from the surface work must be done against the 
forces, i.e. energy must be supplied. The energies of adsorption and 
desorption are usually expressed in heat units and are defined as the 
heat liberated in the adsorption, or required to cause desorption, of 
one mole of the gas. 

In physical adsorption the attracting forces are comparatively weak 
and the heats of adsorption and desorption are small (a maximum of 8 
kcal mol. ~ ■') . In chemisorption the process is similar to the forma- 
tion of chemical compounds and the attractive forces are much larger 
than in the physical case. Heats of chemisorption are correspond- 
ingly higher and can be as large as 250 kcal mol. " ^. Chemisorption 
does not always occur directly from the gaseous state ; molecules may be 
initially adsorbed physically and then, with the provision of a certain 
minimum energy (the heat of activation) they may become chemi- 

10 



SORPTION AND DESORPTION 



sorbed. This is known as activated chemisorption and the desorption 
energy is the sum of the heats of chemisorption and activation. The 
process readily occurs during adsorption at a heated surface and 
experimental evidence'*' suggests that the total amount of gas which 
can be adsorbed in this manner is higher than that by non-activated 
processes. It should be noted that the inert gases carmot be chemi- 
sorbed and are therefore only weakly held on a surface. 

The number of molecules adsorbed on unit area in unit time is 
given by 

where v is the rate at which molecules strike unit area and s is the 
sticking coefficient {i.e. the probability that a molecule which strikes 
the surface will be adsorbed) . 
Substituting eqn 1.11 gives 



^ = sp{27TmkT)-v^ 

= 3-50 X l0^^sp{MT)'^i^mol. cm-^sec-^ 



(1.15) 

(1.15a) 

when p is in torr. 

It is interesting to use eqn 1 . 1 5 to estimate the time required to form a 
complete monolayer upon a previously clean surface. Assuming that 
each atom of the surface can adsorb one gas molecule, that the 
average solid surface has about 8 x 10^* atoms cm~^, and that s = I, 
then at r = 300°K and for nitrogen (M = 28) the time t to form a 
monolayer is given approximately by 



2xl0-« 



sec 



where p is in torr. Thus, except at pressures below 10"^ torr, 
surfaces are always covered by at least a monolayer of gas. 

In addition to adsorption, molecules are desorbing from unit 
area of surface at a rate given by 



d«d 
d^ 



Nof 
t. 



(1.16) 



where Nq is the total number of molecules required to form a com- 
plete monolayer,/is the coverage {i.e. the fraction of possible adsorp- 
tion sites which are actually occupied) and t^ is the average time spent 
by an adsorbed molecule at a particular site and is known as the 
sojourn time. This time is shown by Frenkel'^' to be 



11 



BASIC VACUUM CONCEPTS 



-■-(§) 



(1.17) 



where t' is of the order of 10~^^sec"°' "' and H^ is the heat of 
desorption. 

Eqn 1.16 is only valid for less than a complete monolayer. Similar, 
but more complex, equations may be deduced for multilayer adsorp- 
tion. 

The exponential dependence of t^, and hence of dnjdt, upon both 
H^ and T means that f, varies over a wide range, from about lO"-'-^ to 
IC sec for small values of //<j(2.«. physical adsorption) and low T(77°K), 
and from 10 "^ to lO^^sec for chemisorption {Ha = 10-200 kcal 
mol. ~ ^) at room temperature. 

The equilibrium between adsorption and desorption {i.e. when the 
pressure of the gas in the region of the surface is constant) is found by 
equating eqns 1.15 and 1.16, leading to the number of molecules ad- 
sorbed on unit area of surface as 



Nof =- sp{2TTmkT)-^i^ t. 



(1.18) 



In principle, eqn 1.18 can be used to express Nq/ sls a function of 
p for constant T (the adsorption isotherm) and as a function of T at 
constant p (the adsorption isobar) . Unfortunately, s appears to be a 
function of both/and T; in particular, s falls rapidly as/approaches 
unity and usually decreases with increasing temperature. In prac- 
tice, isotherms are observed experimentally and used to determine s 
and/. 

However, eqn 1.18 may be used to predict some general features: 

(fl) The quantity of gas adsorbed increases with pressure. 

{b) Very little gas can remain physically adsorbed under high 
vacuum conditions at room temperature. 

{c) The temperature variation is dominated by the exponential 
form of ts, and thus at low temperatures the quantities adsorbed 
(even for low heats of adsorption) are considerable. 

The non-equilibrium case in which the desorbed gas is removed 
from the neighbourhood of the surface and the free gas pressure falls, 
is of particular interest in high vacuum and is considered in Section 
1.12.3.2 dealing with the theory of outgassing. 

1.11.2 Absorption and Permeation 

The term absorption refers to gas which diffuses into a solid and 
exists within the solid in a dissolved state. Generally, gases dissolve 
in solids to a concentration e given by 



= bp^-'i 
12 



(1.19) 



OUTGASSING 

where p is the gas pressure, j is the dissociation constant of the gas 
(which is two for diatomic gases in metals and one for all gases in non- 
metals) and b is the solubility of the gas in the solid. 

The diffusion of the gas into and through the solid obeys Pick's laws 
of diffusion which are : 

(a) In the steady state, when the gas concentration is independent 
of time, gas diffuses across a plane of area 1 cm^ in a region where the 
concentration gradient is dejdx at a rate R^ given by 



Ra = -Da 



Ax 



(1.20) 



{b) In the transient case, when the concentration varies with time 

(1.21) 



^^d?~~ 



de 
dt 



In eqns 1.20 and 1.21, D^, is the diffusion coefficient defined by 
eqn 1.20 and measured in cm^ sec"-^. D^ varies with temperature 
according to 



i), = i)oexp(-^^) 



(1.22) 



where H is the activation energy of absorption and Dq is a constant 
for a given gas and material. 

Eqn 1.19 shows that if the gas pressure in the region of a solid is 
reduced then, as a result of diffusion out of the solid, according to 
eqn 1.21, the concentration of dissolved gas falls. If the solid is the 
wall of a vacuum vessel then gas on the high pressure side diffuses 
into the solid to maintain the concentration appropriate to the ex- 
ternal pressure. Eventually the concentration at the low pressure 
surface will have fallen to such an extent that the rate of diffusion 
through this surface is equal to the rate at which gas diffuses into the 
solid at the high pressure surface. Thus, a dynamic equilibrium 
exists in which gas permeates the solid from high to low pressure, the 
permeation rate being given by the solution of eqns 1.19 and 1.20. 

1.12 Outgassing 

1.12.1 Introduction 

When all the free (volume) gas has been removed from a vessel, gas 
continues to appear in the vessel as a result of desorption from, and 
permeation through, the walls. This generation of gas is called out- 
gassing and is measured in terms of the outgassing constant K, which 

13 



i 



BASIC VACUUM CONCEPTS 

is defined as the rate at which gas appears to emanate from unit area 
of surface, and is usually measured in torr 1. sec " ^ cm " ^. Outgassing 
is the principal cause of difficulty in obtaining pressures below 10"^ 
torr, and in order to achieve very low pressures action must be taken 
to reduce it. 

1.12.2 Experimental Observations 

A large number of experimental measurements of outgassing 
constants have been carried out and several interpretations of the 
results have been made (see for example Blears, Greer and Night- 
ingale'^^', Beckmann'13)^ Power and Crawley'"', Boulassier'^^', 
Crawley and de Csernatony'^^', Santeler'^'"). In general, the 
observations can be represented by an empirical equation of the form 

K>, = K,+K,t;:y (1.23) 

where K^ and K^ are the outgassing rates at h hours and one hour 
respectively after the start of pumping; t^ is the time in hours after the 
start of pumping; K^ is the limiting value of K^ and is generally 
negligible unless 4 is very large; y is a number which varies with 
time of pumping. 

At the commencement of pumping y is large and outgassing rates 
fall very rapidly, but after a few minutes the fall becomes less marked 
with values of y lying between 0-5 and 2, depending upon the material ; 
thereafter y changes slowly with time. For metals, the values of 
K-^ typically lie in the range 10"''-10"^ torr 1. sec~^ cm~^ with values 
of y near 1 . For non-metals values of K^ are somewhat higher with 
y lying between 0-5 and 1. Values of y greater than 1 are usually 
associated with an unusual surface condition, such as rusty steel. 

After a long pumping time (rarely less than 10 h) outgassing rates 
show a tendency to fall exponentially with time until limited at K^. 
A curve showing a typical time variation of outgassing rate is given in 
Figure 1.1. 

Of the gases whose liberation has been observed during prolonged 
pumping at room temperature, water vapour is the main gas desorbed 
initially, its outgassing rate usually varying as f"-^, followed by smaller 
quantities of hydrogen, carbon monoxide, carbon dioxide, and hydro- 
carbons. The water vapour content is chiefly due to sorption from 
the atmosphere, whereas the appearance of other gases is believed to 
be mainly due to gas absorbed during the preparation of the material 
[e.g. hydrogen in aluminium, and carbon compounds in mild steel). 

If the temperature of the system is raised the outgassing rate rises 
rapidly to a peak value followed by a slower fall back to a ^^ ' varia- 
tion, but at a higher level corresponding to the elevated temperature. 

14 



outgassing 

If, after a sufficiently long time, the temperature is allowed to fall to 
its original value the outgassing rate falls rapidly to a level which is 
significantly lower than that which would have existed if pumping had 
been at the lower temperature throughout. A typical variation of 
outgassing rate when a baking cycle is incorporated is shown in 
Figure 1.1 where it may be compared with constant temperature 
behaviour. 



Behaviour at 
"^ ^ room temperature 

Behaviour when baking 
-^ — X — '^- cycle incorporated 




Figure 1.1. Typical variation of outgassing constant with time 
of pumping 



Baking may also have the effect of causing activated chemisorption 
of physically adsorbed gas (in particular, water vapour) which can 
then be desorbed only by prolonged heating at much higher tempera- 
tures. Chemisorbed water vapour continues to be evolved at tempe- 
ratures in excess of 300°C<^*'. It would therefore appear that a 
degassing programme should begin with pumping at room tempera- 
ture to remove physically adsorbed water vapour, before baking is 
commenced. 

Preliminary treatment of the surface also has a marked effect upon 
outgassing rates. Cleaning with detergents and organic solvents 
obviously removes gross surface contamination, although recent 
work by Adler'^^' suggests that organic 'cleaning' solvents tend 
to be strongly sorbed and are not easily removed during pumping. 
Polishing reduces the amount of surface oxide, but it should be 
noted that such layers reform very rapidly. The removal of surface 

15 



BASIC VACUUM CONCEPTS 

imperfections by polishing also reduces the surface area available 
for subsequent sorption. 

Table 1.2. Values of Outgassing Constants 





Kr X W 




Kt X W 




Material 


(torr 1. sec ' 
cm-") 


Vi 


(torr 1. sec"' 
cm-») 


y« 


Stainless steel, polished and 






0-05 
0-4 


1-6 
0-8 


degreased 
Stain ess steel 


1-75 


M 




0-9 


0-7 


0-3 


0-7 


Mild steel 


5 


1 


1-4 


1 


Mild steel (slightly rusty) 


6 


3 


0-3 


1-5 
1-4 


Mild steel (rusty) 


44 


1-4 


5-5 


Mild steel (degreased) 


3-6 


— 


3 




Aluminium (anodized) 


10 


0-7 


' 


Aluminium (washed in Stergene) 


1 


— 


— 




Copper 
Brass (cast) 


23 
10 


1 
1 


2-5 


1 


Porcelain (glazed) 

Steatite 

Polythene 

Perspex 

Araldite 


6-5 
0-9 
200 
40 
18 


0-5 

1 

1-6 

0-9 

0-4 


3 

0'2 
20 
14 
10 


0-5 

1 

1-6 

0-6 

0-5 


Nylon 
Silicone rubber 


120 
90 


0-5 
0-75 


60 
30 


0-5 
0-8 


Silicone rubber (degassed followed 










by 24 h dry Na) 
Neoprene 
Neoprene (degassed followed by 


13 
300 


0-5 

0-4 


180 


0-4 






0-5 
6 




24 h dry Na) 


120 


' 


' 


Nitrile O-ring 


10 





' ■ 


Nitrile O-ring after baking 4 h at 
100°C 


10 





0-7 


_ 


Viton O-ring 


. 10 





3 





Viton O-ring after baking 4 h at 






0-01 




100°G 


1 


. 




Viton O-ring after bakmg 4 h at 










150°G 


0-1 




0-003 





Details of observed outgassing rates and the values of y after 1 h 
and 4 h are given in Table 1.2. In using this data with eqn 1.23 to 
estimate values of JC at times less than 1 h, it is recommended that the 
value of 7i be used in the extrapolation although this will tend to give 
results for K which are somewhat low. 

1.12.3 Theory of Outgassing 

The following brief account of outgassing theory, based largely on 
the work of Dayton<"'2°'^", jg intended to give only an indication 
of the lines upon which work in this field is proceeding. 

16 



OUTGASSING 

Outgassing takes place as a result of the reversal of both absorption 
and adsorption processes. Thus, a molecule which diffuses from the 
interior of a soUd may be held at the surface by adsorption, and con- 
sequently a complete theory must take both processes into account 
simultaneously. However, in most cases the rate of diffusion is so 
small compared with that of desorption of adsorbed gas that the two 
processes may be analysed separately and the resulting outgassing 
rates subsequently added. The appHcation of such analyses to 
practical situations is difficult since usually very httle is known about 
the types and quantities of gases initially sorbed, nor the proportions 
in which they are adsorbed and absorbed. However, the theory 
does give a general indication of the orders of magnitude of outgassing 
rates which may be expected and of their time variation. 

1.12.3.1 Diffusion Controlled Outgassing — Considering outgassing which 
is governed only by diffusion of gas from the interior of a solid {i.e. the 
reversal of the absorption process) the outgassing rate of a given gas 
from a given material can be obtained from solutions of eqns 1.20 and 
1.21 with appropriate boundary conditions. In general the solution 
consists of the sum of an infinite series, but may be approximated to 
give the outgassing rate from the wall of a vessel of thickness Lq cm as 



Kh = K^ + 



where 



and 






= il"--^[l 



-exp 



m 



/n = 



Trig 



5-76xlO*Z)^ 



(1.24) 



(1.24a) 



(1.24i) 



When t^ < <o/4, tj, = tl'^ and thus X varies initially as tj^^'^ but eventu- 
ally falls more rapidly to approach an exponential dependence as 
t^ becomes large. The theoretical values of: K^^ and K^ are given by 



^1 = 



2-79x10- 

(3,600) i-i 



■■'(T 



(1.25) 



where yi is the value of y when t^ = 1, and e^ is the gas concentration 
when tf, = 0, measured in cm^ (at s.t.p.) cm'^ of material, and 



K„ = 2-79x10 



-3 T^i, 

^ T PO 

^0 



(1.26) 



The product D^b is known as the permeation constant P, and is 

2 I.H.V.T. 17 



I 



BASIC VACUUM CONCEPTS 

measured in cm^ (at s.t.p.) cm~^ of cross-section for a thickness of 
1 cm and pressure differential of 1 torr. The pressure /o is the partial 
pressure outside the enclosure of the gas considered. 

As an example, consider the outgassing of hydrogen from a steel 
vessel of wall thickness 1 cm at 27°C. The following values for the 
various parameters are assumed 

Da = 5xl0-^cm2sec-i 

€0 = 0-1 cm^ (at s.t.p.) cm"^ 

b = 10 ~^ cm^ (at s.t.p.) cm~^ torr~^ 

J = 2 

po = 4x 10~* torr (partial pressure of hydrogen in the atmos- 
phere) . 

Thus to is approximately 10* h and y = ^ up to about 2,500 h. 
Substituting in eqns 1.25 and 1.26 gives 

K^ = 5-6 X 10"^ torr 1. sec"^ cm"^ 
and 

K^ = 2x 10-^3 torr 1. sec"^ cm"^ 

In this particular case the permeation outgassing rate will almost 
certainly be greater than the value oiK^ calculated above because of 
the liberation of hydrogen at the outer surface of the vessel by the 
action of water vapour on iron. Dayton'-"^^' estimates that K^ from 
this source is about 5x10"^^ torr 1. sec~^ cm"^ for a wall thickness of 

1 cm. The values of K^^ due to other gases are less than this. 
Since the diffusion coefHcient of hydrogen is greater than that for 

most other gases, the value of K^ for hydrogen will be the largest 
single factor in the total diffusion controlled outgassing rate . However, 
(f) the calculated value of K^ for hydrogen is some ten times smaller 
than the values observed in practice, and {ii) the predicted value of 
Yi is ^ whereas it is experimentally observed to be in the region of 1 . 
Thus, it must be concluded that factors other than diffusion of gases 
from the interior play a considerable part in the outgassing of metals. 
The outgassing due to water vapour is believed to be the main addi- 
tional factor. 

Similar calculations can be carried out for non-metals leading, 
for example, to values of X^ for most elastomers of the order of 

2 X 10"'' torr 1. sec"-^ cm~^ for dry air and 10~® torr 1. sec"-' cm~^ 
for water vapour. The latter figure and the predicted variation as 
t~^'^, are in reasonable agreement with experimental data. 

18 



OUTGASSING 

1.12.3.2 Outgassing Controlled by De-adsorption — ^The theoretical ana- 
lysis of this process requires the solution of the non-equilibrium case of 
adsorption referred to in Section 1.11.1. It involves the rate of 
change of coverage as a function of the free gas pressure. 

Assuming that the surface has less than a monolayer coverage and 
that V dp I At (where Fis the volume of the vessel) is very small then 
the outgassing rate at time t due to desorption of adsorbed gas may be 
expressed approximately by 



Kt = 



lO-^xT/o 



exp 



(4) 



(1.27) 



where /o is coverage when / = 0, and t and 4 are in consistent units. 

Putting /o = 1 (its maximum value) it will be seen that for small 
values of ^s corresponding to physical adsorption, the initial outgassing 
rate is very high but falls extremely rapidly with increasing time. 
Even for moderately strong chemisorption the outgassing rate falls 
fairly rapidly. For instance at room temperature with t, = lOmin 
(corresponding to H^ = 20 kcal mol. " ^) the outgassing rate after 
only 5 min falls to about 2x10"^ torr 1. sec"-"^ cm~^, and is less than 
10"^ torr 1. sec"-' cm"^ after 1 h. On the other hand, for strong 
chemisorption the initial outgassing rate is low but falls only very 
slowly with time. 

Thus, the high initial outgassing rates which are experimentally 
observed are almost certainly due to weakly adsorbed gas, and the 
desorption of more strongly adsorbed gas will add to the gas originat- 
ing from de-absorption. 

1.12.3.3 Outgassing of Water Vapour from Metal — ^The theory of the 
previous section is unlikely to account for the observed outgassing of 
water vapour, since it assumes less than a monolayer initial coverage, 
whereas surfaces exposed to a high humidity can adsorb several 
monolayers. Several workers have investigated this problem. 

Kraus*^^' develops a simplified form of the adsorption isotherm 
and assumes that up to ten monolayers of vapour may be adsorbed. 
From this he shows that the outgassing rate varies roughly as /"-^ over 
limited periods of time. However, in order to account for the ob- 
served quantities of water vapour it is necessary to assume surface 
roughness factors of about 100, which is much greater than those 
normally observed. 

Dayton*^"^ suggests that water vapour is absorbed in the pores of 
the layer of oxide that is inevitably present on the surface of most 
metals. A semi-empirical analysis of the distribution of pore size and 

19 



1 



BASIC VACUUM CONCEPTS 

layer thickness leads to an expression for the outgassing rate which 
varies as t~^. 

RoGERS<^^> has made a theoretical investigation of the sorption and 
subsequent desorption of water vapour by a previously thoroughly 
degassed metal. He assumes that the water vapour is absorbed into a 
thin layer of metal near the surface and shows that subsequent out- 
gassing depends upon the time of exposure to a moist atmosphere. 
He further shows that initially the outgassing rate varies as r^'^ but 
the fall gradually becomes steeper until, when the pumping time is 
long compared with the exposure time, the outgassing rate varies as 

This case is a common one in practice and the theoretically 
predicted time variation of outgassing rate agrees in general form with 
the experimental variation*^*'. 

1.12.3.4 Theory of Temperature Variation of Outgassing Rate — Dayton*^" 
has investigated the effect of baking, followed by subsequent room 
temperature pumping, assuming that the outgassing is due to the 
reversal of absorption. He shows that if the temperature is suddenly 
raised from T^ to T^ at time t^ after commencement of pumping then 
the outgassing rate after a total pumping time t^ is given by 



^. = 






\^{t,.-t,)+t)^ 



[k > h) 



(1.28) 



where K^ is the outgassing rate that would have existed att^ = 1 if the 
temperature had not been raised, and Da and Dj, are the diffusion 
coefficients of the gas through the solid at temperatures T^ and T^ 
respectively. The condition that t^ < <o/4 (see eqn 1.24) is assumed 
to be true. It will be seen that at the instant of raising the tempera- 
ture {t^ = t^) the outgassing rate is suddenly increased in the ratio 
DJDa which is usually » 1. The outgassing rate then falls rapidly 
but as t^ becomes large compared to 4, it approaches a value given by 






If the temperature is then suddenly returned to T^ when t^ 
subsequent outgassing rate is given by 

^1 



^. = 



k + 



(w) ('•-« 



1/2 



(1.28a) 



t., the 



(1.29) 



20 



OTHER SURFACE EFFECTS 



i 



It can be seen that at the instant of reducing the temperature {t^ = t^) 
the outgassing suddenly falls to 



^" - {D.jDaYiHy^ 



(1.29a) 



If baking had not been carried out, the outgassing rate K^ would only 
have been achieved after a time 4 given by 



Therefore 



K — ^ 

"■0 ~ fll2 



t -^t 

to - J^^ k 



(1.30) 



i.e. the time required to reach a given outgassing rate has been 
reduced by a factor DJD,,. 

This theory, of course, applies only to the outgassing of a single gas. 
In practice, the overall outgassing behaviour would be given by the 
summation of a set of equations having the form of eqns 1.28 and 1.29. 
In addition, the theory does not take into account the effect of water 
vapour and it assumes a variation as /~^'^ rather than the observed 
<"i. The latter point is borne out by the experimental work of 
Das"^* which, although showing a change of outgassing rate at a 
temperature change which is roughly in agreement with the theory, 
shows a general variation as t~^. 

1.13 Other Surface Effects 

1.13.1 Gettering 

This is a process in which a solid is evaporated in a gaseous atmos- 
phere. While in the vapour phase the material has a high reactivity 
for both physical and chemical reactions with the gas, such reactions 
being referred to as dispersal gettering. The vapour and its reaction 
products eventually condense upon relatively cool surfaces and pre- 
sent a very clean surface capable of sorbing free gas, the sorbing 
process being referred to as contact gettering. So long as the evapora- 
tion proceeds, the contact gettering surface is continually replaced, the 
newly condensed material making the desorption of gas from the 
previous layer very difficult. 

Both dispersal and contact gettering effectively remove gas from the 
system and so constitute a pumping action. It should be noted that 
inert gases can only be gettered by the contact effect. 

The properties of some materials commonly used as getters are 

21 



fiASIC VACUUM COJSiCEPTS 

given in Table 1.3. The dispersal effect is expressed in terms of the 
theoretical quantities of oxygen, nitrogen and hydrogen which react 
with 1 g of material. The contact effect is expressed in terms of the 
experimentally observed quantity of gas sorbed per gramme of 
deposited material. In the data for contact effect the smaller quan- 
tity is for a bright deposit (obtained at low pressures) whilst the 
larger quantity is for a diffuse deposit (obtained at relatively high 
pressures) . 

Table 1.3. Gettering Effects of Various Materials 
{after Dushman«« and Holland<">) 



Material 


Dispersal effect 
(torrl.g-i) 


Contact effect at 20°C 
(torrhg-i) 


Aluminium 


Oa 517 
Na 345 
Ha 


7-39 


Magnesium 


Oa 382 

Na 255 
Ha 764 


20-200 


Misch Metal (Ce/La) 


La Ce 
Oa 100 133 
Na 66 
Ha 200 265 
COa 


21-51 
3-16 

46-64 
2-45 


Titanium 


Oa 388 
Na 194 
Ha 388 
COa 


4-3 

1-9-2-5 

3-4-42 


Zirconium 


Oa 204 
Na 102 
Ha 204 





1.13.2 Ion Effects 

When ions bombard a surface some of their kinetic energy is 
transferred to the molecules lying near the surface of the solid. The 
energy so transferred may be sufficient to provide the desorption 
energy necessary to release adsorbed gas and thus ionic bombardment 
may be used to clean a contaminated surface. 

If the bombardment is due to heavy positive ions being attracted to 
a negatively charged cathode the collision energies involved may be 
sufficient to knock out atoms of the cathode. Such a process is 
known as cathode sputtering, and is often used to cause the evapora- 
tion of getter materials. 

22 



REFERENCES 



REFERENCES 



Glossary of Terms Used in High Vacuum Teck- 



McGraw-Hill, New York and 
McGraw-Hill, New York and 
McGraw-Hill, New York and 
McGraw-Hill, New York and 
McGraw-Hill, New York and 



1 British Standard 2951: 1958. 

nology 
^ Kaye, G. W. C. and Laby, T. H. Tables of Physical and Chemical Constants, p 9. 

Longmans Green, London, 1959 
^ Kennard, E. H. Kinetic Theory of Gases, p 7. 

London, 1938 
* Kennard, E. H. Kinetic Theory of Gases, p 63. 

London, 1938 
^ Kennard, E. H. Kinetic Theory of Gases, p 1 10. 

London, 1938 
^ Kennard, E. H. Kinetic Theory of Gases, p 138. 

London, 1938 
' Kennard, E. H. Kinetic Theory of Gases, p 163. 

London, 1938 
' Hayashi, C. Vacuum Symp. Trans. (1958) 13 
» Frenkel, J. Z. Phys. 26 (1924) 117 

'■" Trapnell, B. M. W. and Hayward, D. O. Chemisorption, p 150. Butter- 
worths, London, 1964 
" Dayton, B. B. Vacuum Symp. Trans. (1960) 101 
'" Blears, J., Greer, E. J. and Nightingale, J. Trans. 1st Int. Congr. Vacuum 

Techqs, Namur, 1958. Pergamon, Oxford, 1960 
13 Beckmann, W. Vacuum 13 (1963) 349 
'■' Power, B. D. and Crawley, D. J. Trans. 1st Int. Congr. Vacuum Techqs, 

Namur, 1958. Pergamon, Oxford, 1960 
i» BouLASsiER, J. C. Vide 14 (1959) 39 

'^ Crawley, D.J. and de Csernatony L. Vacuum 14 (1964) 7 
1' Santeler, D.J. Vacuum Symp. Trans. (1958) 1 
'« Flecken, F. a. and Noller, H. G. Vacuum Symp. Trans. (1962) 58 
i*" Adler, J. J. Vacuum Sci. Technol. 2 (1965) 209 
2" Dayton, B. B. Vacuum Symp. Trans. (1962) 42 
^1 Dayton, B. B. Vacuum Symp. Trans. (1963) 293 
=2 Kraus, T. Vacuum Symp. Trans. (1964) 77 
'"^ Rogers, K. W. Vacuum Symp. Trans. (1964) 84 
" Dayton, B. B. 12th Natn. Symp. Am. vacuum Soc. (1965) Paper 12.1 
'^ Das, D. K. Outgassing Characteristics of Various Materials in an Ultra-high 

Vacuum Environment. AEDC Report No. AEDC-TDR-62-19, 1962 
'^ Dushman, S. Scien'ific Foundations of Vacuum Technique, p 664. Wiley, New 

York, 1949 
='' Holland, L. J. Scient. Instrum. 36 (1959) 109 



23 



THEORY OF GASEOUS FLOW 



2.1 Introduction 

In the production and maintenance of a vacuum, gas is caused to 
flow from a vessel through pipelines and pumps and then into the 
atmosphere or a reservoir. The mechanisms of the flow of the gases 
and the factors which influence flow are matters of great importance 
to the vacuum technologist and the present chapter is concerned 
with the study of this topic. 



2.2 Impedance 

A pressure difference occurring in a vacuum system causes gas to 
flow from the region at high pressure to that at low pressure. The 
rate of gas flow depends on the impedance Z {i.e. resistance to flow) 
of the intervening system. The impedance is defined as the ratio of 
the pressure difference to the gas throughput 



Z = 



P1-P2 



(2.1) 



2.3 Conductance 



Instead of using impedances, it is usually more convenient to deal with 
the conductances of vacuum components. The conductance C is 
defined as the reciprocal of the impedance and substituting this in 
eqn 2.1 leads to 



g = c{pi-p2 



(2.2) 



It should be noted that the units of conductance and speed are the 
same (1. sec'i), but the two concepts are physically different and 
should not be confused. 

It is readily seen that the rules for summing conductances are as 
follows. 

24 



effect on pumping speed of conductance 
Conductances in series : 

^ 1^1 <-/2 O3 

Conductances in parallel : 

C = Ci+C2 + C3+... (2.4) 

All components in a vacuum system {e.g. pumps, pipelines, traps, 
baffles) have their own conductances and are discussed later in this 
chapter. 



2.4 Effect on Pumping Speed of a Component 
OF Conductance C 

Consider a pump operating at a speed Si at a pressure />! and let it be 
connected by a component of conductance C to a vessel in which the 
pressure is p2. Then, the throughput is given by eqn 1 .5 as 



q = Sipi = Sep2 



(2.5) 



where S^ is the effective speed in the vessel. Eliminating p^ and p^, 
and using eqn 2.2 gives 



- itB 



1-1 i 



(2.6) 



^e 



Si+C 



(2.7) 



Thus Sg is always less than Si. 

If C » Si, Sg = Si and the effect of the component is negligible, 
Sg being determined by the pump. 

If C «: Si, Sg = C and the eflfective speed is numerically equal to 
the conductance of the component, i.e. S^ is independent of the pump 
speed. In practice, the situation generally obtaining is somewhere 
between these extreme cases. 



2.5 Effective Speed in a Vessel due to Several Pumps 

Consider a number of pumps connected to one vessel. Let the speed 
of the fth pump be Si, the pressure at its inlet be/),; let it be connected 

25 



THEORY OF GASEOUS FLOW 

to the vessel by a pipeline of conductance Q and let its throughput be 
qi. The effective speed at the vessel is S^, the pressure p^ and the 
total throughput q^. 
Then 

= SePe 



and 



Thus 



?i = Sipi 

qi = C^{p-p^) 



2 



Ci2?i 






(2.8) 



In the special case where all the pumps have the same speed S, the 
same throughput q, and the pipelines have the same conductance C, 
it follows that the pump inlet pressures are the same, say p. 
Under these conditions, eqn 2.8 becomes 

o - ^vC-^ 



S,.= 



ySC 
S+C 



(2.9) 



where y is the number of pumps. That is, the total effective speed is 
equal to the product of the number of pumps and the effective speed 
of a single pump. 



2.6 Mechanisms of Gas Flow 

In order to be able to calculate the conductances of various 
components, the different mechanisms of gas flow must be considered. 
The simplest component for study is the straight pipeline of uniform 
circular cross-section and this is discussed in the first instance. 

There are three mechanisms by which gas flow can occur, namely 
(f) turbulent flow, (ti) viscous flow, or {Hi) molecular flow. The 

26 



MECHANISMS OF GAS FLOW 

important criteria for distinguishing between them are the gas pres- 
sure and the pipe diameter; the influence and importance of these 
are discussed in the following sections. 

2. 6. 1 Turbulent Flow 

With a high pressure difference (for example 1 ,000 torr) across a small 
diameter pipe the velocity of flow of the gas is large but not constant. 
In fact, eddies are formed randomly throughout the tube. To 
produce even approximate equations for this type of flow is beyond the 
scope of this book, and 'fortunately the pressures used in vacuimi 
work are generally low enough to ensure that it does not occur. It is 
sufficient to state that turbulent flow occurs only if the throughput 
q > 2x Iff'Z) torr 1. sec~^, where D is the pipe diameter in centi- 
metres'^*. This only applies when commencing evacuation from 
atmospheric pressure with very large capacity pumps. 

2.6.2 Viscous Flow 

The concept of gaseous viscosity is discussed in Section 1.7, and it is 
seen that for this type of flow the pressure difference across the tube 
must be such that the mean free path is very much less than the tube 
diameter. From the definition of the coefficient of viscosity it can be 
shown {e.g. by following Newman and Searle*^') that for a gas of 
viscosity ij the viscous conductance C„ of a circular pipe of length L 
cm and diameter D cm is given by 



C„ = 



128 Lr,^ 



(2.10) 



where p = ^(/>i +P2) = the mean gas pressure. 
For air at 20°C and p in torr 

C„ = 18^-1. sec- 



{2.10a) 



2.6.3 Molecular Flow 

The assumption made in the viscous flow equation that the mean 
free path of the gas molecules is small compared with the dimensions 
of the tube is no longer valid as the mean pressure is reduced. The 
number of intermolecular collisions decreases and the molecules make 
more collisions with the walls of the tube, so that the concept of 
laminar flow breaks down. The factor determining the flow behaviour 
now becomes the diffuse reflection of the gas molecules at the walls of 

27 



THEORY OF GASEOUS FLOW 



the tube. This type of flow is known as ' free molecule ' or ' molecular ' 
and it can be shown'^' that the conductance C^ of the tube is then 
given by 



For air at 20°C, 



- 1 /MT^ 
~ 6S m L 

_ 77 I'RT D3 
" 3V 2-rTM L 



C„ = 12-i:^l.'sec-i 



(2.11) 



(2.1 la) 



It will be noted that, unlike the viscous conductance, the molecular 
conductance is independent of pressure. 

2.6.4 Transitional Flow 

The boundary between viscous and molecular flow is not a sharp 
one and the behaviour of the gas in the transitional region is usually 
treated on a semi-empirical basis. The formula most commonly used 
in vacuum work is that due to Knudsen'^*, who shows that the 
equations for viscous and molecular flow can be combined to give a 
total conductance Cr 



Cr — C„+C_ 



1 + 



7; 



m Dp 



Thus the total conductance for air at 20°C is given by 



(2.12) 



(2.13) 



where D and L are measured in cm and p in torr. When Dp is large, 
the expression (1 +256Z)^)/(1 +316Z)^) tends to a constant value of 
0-81 and the molecular flow term is then negligible compared with 
that due to viscous flow. The viscous conductance is then equal to 



C = 



182Z)*^ 372^2^ 



1. sec" 



(2.14) 



where A is the cross-sectional area in cm^, which equals \ttD'^ for a 
circular pipe. 

At the other extreme, when Dp is small the viscous term is negligible 

28 



MECHANISMS OF GAS FLOW 



and the term (1 +2561)^) /(I -|- 3161)^) becomes equal to unity. 
Hence the molecular conductance becomes 



^ 12-1Z)3 60^% 
where U is the perimeter of the pipe in cm, in this case ttD. 



(2.15) 



o Q 




10"' 10° 10' 10^ 



Dp, torr cm 
Figure 2.1. Graphical representation of eqn 2.13 



The variation of conductivity with Dp is illustrated in Figure 2. 1, 
in which C^^LID^ is plotted against Dp. From this graph the limits of 
the various types of flow can be seen. Thus HDp > 5 x lO""-"^ torr cm 
the flow is purely viscous, whereas ii Dp < 1-5 x 10"^ torr cm, purely 
molecular flow takes place. In the intermediate region the flow is 
transitional requiring the use of the full conductance eqn 2.12. 

2.6.5 The Effects of Temperature and the Nature of the Gas 
Eqns 2.14 and 2.15 were obtained for the case of air at 20°C. For 

other gases and for temperatures different from 20°C appropriate 

modifications must be made. 

In molecular flow the conductance is proportional to VTJM 

(see eqn 2.11); hence the general equation for the molecular 

29 



THEORY OF GASEOUS FLOW 



conductance C„ of a pipe containing a gas of molecular weight M 
at a temperature of d°G is given by 



C„ 



■J 



12-1D3 /( e + 273) 29 
293 M 



(2.16) 



The conductance of a pipeline for viscous flow is given by eqn 2.10, 
in which the viscosity is the only variable with temperature and the 
nature of the gas. Table 2. 1 gives values of the viscosities of some gases 
at various temperatures, and if T]g is the viscosity of a gas at tempera- 
ture 9, the viscous conductance of the pipeline for this gas and 
temperature is given by 



C„ = 3-28x10- 



Z,7J, 



1. sec ^ 



(2.17) 



where p is the mean pressure (in torr) at the temperature of the 
system. 

Table 2.1. Gaseous Viscosities {in poise X 10^) 











Gas 








Temperature 
















(°C) 


Air 


A 


CO. 


He 


H. 


N. 


o. 


H.O 





17 


21 


14 


18.6 


8-4 


17 


19 


_ 


20 


18 


22 


15 


19-4 


8-8 


18 


21 


10 


50 


20 


24 


16 


21 


9-3 


19 


22 


— 


100 


22 


27 


19 


23 1 10-3 


21 


24 


13 



2.7 Conductances of Other Components 

The conductance of vacuum system components other than the circu- 
lar pipe may also be calculated from first principles. The work of 
Barrett and Bosanquet**' in this iield has yielded important data, 
and reference may also be made to Dushman*^^; this section is devoted 
to a summary of the more important of these results. 

The data presented refers only to molecular flow conductances for 
air at 20°C, for which the symbol C„ is used. The calculation of the 
conductances for other gases and temperatures can be made using 
eqn 2.16. 

2.7.1 Orifice 

The conductance of a hole in a flat plate of area large compared 
with that of the hole can be determined as follows. If the pressures 

30 



conductances of other components 

of the gas on either side of the hole arej^^ and/ig, then from eqn 1.12a 
the net throughput of g^s through the hole is 



? = 3 -64^ l^^j (/>! -/.a) torr 1. sec " i 



(2.18) 



where A is the area of the orifice in cm^. Hence 



C„ = 



Pl-P2 



= 3-64A 



Q" 



1. 



In the case of air at 20°C, 

C„ = 11-6^ = 9-l£)2 = ^xl2-lDM. sec-i 



(2.19) 



(2.20) 



2.7.2 Straight Pipe of Finite Length 

In the derivation of the expression for the conductance of a straight 
pipe of circular cross-section no account was taken of the impedance 
of the entrance orifice of the pipe. In fact, if the pipe is connected into 
a vessel of dimensions large compared with the pipe diameter, gas 
molecules in the vessel have to 'find' the entrance aperture of the 
tube and this constitutes an additional impedance to the gas flow. 

The resultant total conductance is shown by Clausing*^' to be of 
the form 



Z)3 
Cm = 12-1 -J- al. sec" 



(2.21) 



where a is a numerical factor dependent on L and D. Clausing 
gives numerical values of a for a wide range of L and D. Guthrie 
and Wakerling"' give an approximate expression for a, which is 
accurate to 1 -5 per cent, as 

15(Z,/Z») + 12(Z,/Z))2 



20 + 38(L/i)) + 12(Z,/Z))= 



When LjD is large, a tends to unity and eqn 2.21 becomes identical 
with eqn 2.15. 

An alternative, but not strictly valid, method of approaching the 
problem (see Pirani and Yarwood'^') is to consider the total im- 
pedance of the tube as the sum of the terms due to the pipe and the 

31 



THEORY OF GASEOUS FLOW 

orifice. The combined conductance C„ for a circular pipe is then 
given by the equation. 






i + o- 



1 



I.e. 



which may be expressed as 



12-1Z)3 3 12-1Z)2 
12-1Z)3 



1. sec"'^ 



12-1Z)3 



jSl. sec"^ 



(2.22) 



where the numerical factor j8 is given by 

1 



^ = 



1+f (£>/!) 
Again when LfD is large, ^ tends to unity. 



M 
1-0 
0-9 

I 0-7 
c 0-6 
B 0-5 
^ 0-4 
<3 0-3 
0-2 

0-1 




~77 



2 



/> 



i^ 



TT 



^: 



-7^ 



// 



,^ 



? 



,^ 



Uncorrected 

/3 



10" 



10-' 



10" 



10^ 



10-^ 



10^ 



Figure 2.2. Correction factor for molecular conductance of a tube of 
finite length 

A comparison of the expressions for the conductance as given by 
eqns 2.15, 2.21, and 2.22 is shown in Figure 2.2 where the numerical 
factors (1, a and (8) are plotted against LjD. It can be seen that only 
when LjD > 100 is the simple eqn 2.15 reasonably valid. 

a and |3 are in good agreement, the maximum discrepancy (10 per 
cent) occurring when LjD = 1 . Hence, for most practical purposes 
eqn 2.22 is quite adequate. 

32 



CONDUCTANCES OF OTHER COMPONENTS 



0-125 025 0'5 1 



Length i, ft. 

2 3 4 6 810 20 30 50 100 200 300 

■ I I I I I- 




2 3 4 6 10' 2 3 4 6 10' 



2 3 A 6 10' 2 3 4 6 10* 



Length L, cm 

Figure 2.3. Molecular conductance of a round pipe {based on eqn 2.22) {from Ward 
arul Bunn'°', by courtesy of Engineering Materials and Design Association) 



3 I.H.V.T. 



33 



THEORY OF GASEOUS FLOW 

Figure 2.3 shows a family of curves of molecular conductance, 
for various lengths of round pipe of commonly used diameters, based 
oneqn2.22. 

2.7.3 Annular Orifice 

The conductance of an annulus for molecular flow can be calcu- 
lated from the formula 



C„ = n-6AGj, 1. sec-i 



(2.23) 



where A is the area of the annulus in cm^, and equals iniDl-Df); 
Gi is a numerical factor depending on the diameter ratio, D^fD^. 
Values of Gi for different values of Dj/Dj are given in Table 2.2. 

Table 2.2 Values of GxJor an Annular Orifice 



njD^ 





0-25 


0-5 


0-7 


0-87 


0-97 


Gi 


I 


1-07 


M5 


1-25 


1-43 


1-68 



2.7.4 Concentric Cylinders 

The conductance of the annular tube formed between concentric 
cylinders is given by the equation 

60^2 



C„ = 



UL 



Gil. sec- 1 



which is a modification of eqn 2.15 for a circular pipe. In this case, 
the perimeter U is 

The total conductance of a pair of concentric tubes is given by taking 
into account the entrance annular orifice. Thus 

1 UL I 



60^2(5^ ' ll-6^Gi 



r - 696^^Gi _i 



(2.24) 



2.7.5 Rectangular Duct 

The conductance for molecular flow for a rectangular duct may also 
be expressed in terms of its area of cross-section A and perimeter U. 

34 



ALTERNATIVE CONDUCTANCE CALCULATIONS 

The result is again similar to that for a circular tube but involves 
another shape factor G2. Thus 

Cm = -f^G2l. sec-i 

G2 depends on the ratio of a to b, where a and b are the lengths of the 
long and short sides of the rectangle respectively; a number of values 
of G2 are given in Table 2.3. 

Table 2. 3. Values of G2 for a Rectangular Duct 



a 


1 


i 


i 


i 


X 

5 


i 


^ 


G2 


MO8 


M26 


1-151 


1-198 


1-297 


1-400 


1-444 



Again, it is necessary to consider the effect of the entrance orifice 
on the total conductance of the duct, leading to an equation for C as 
follows 



G„ = 



696^^G2 
60A + 9-IUL 



1. 



(2.25) 



2.7.6 Right-angled Bends 

A common feature in a vacuum system is a right-angled bend which 
introduces extra impedance to flow. It is useful to regard the bend as a 
straight pipe of length L' = L + L^, where L is the pipe length measured 
along the centre line of the bend and L^ is a correction whose value 
depends upon the form of the bend. The maximimi value of L^ is 
generally taken as jD and then the minimum resultant conductance is 



^m 



12-1Z)3 
L + ^D 



1. sec" 



(2.26) 



2.8 Alternative Conductance Calculations 

An alternative method of calculating molecular conductance is 
described by Davis*"'. In this treatment the so-called Monte Carlo 
method is used to consider the individual behaviour of each of a large 
number of molecules which flow through the component. The overall 
effect is expressed in terms of the net number of molecules per unit 
time passing from the entrance to the exit of the component, from 
which the conductance is readily calculated. In principle the method 
can be applied to components of any geometry. For example, 

35 



THEORY OF GASEOUS FLOW 

Levenson, Milleron and Davis'^^' calculated the conductance of 
right-angled bends, chevron baffles and circular plate baffles; Smith 
and Lewin*^^' used the method to compute the effect due to molecules 
'sticking' to the walls of a circular pipe instead of 'reflecting' in- 
stantaneously. The sticking effect appears as extra impedance to 
flow and even with low sticking probabilities the reduction in con- 
ductance is considerable. For example, the conductance of a pipe 
for which LjD = 5 and the sticking probabihty is 0-2, is the same as 
for a pipe for which LjD = 50 and the sticking probability is zero. 

REFERENCES 

» 

1 Guthrie, A. and Wakerlino, R. K. Vacuum Equipment and Techniques, p 25. 

McGraw-Hill, New York and London, 1949 
= Newman, F. H. and Searle, V. H. L. The General Properties of Matter, p 228. 

Arnold, London, 1951 
3 Knudsen, M. Ann. Phys. 28 (1909) 75 
* Barrett, A. S. D. and Bosanquet, C. H. Resistance of Ducts to Molecular Flow. 

ICI (Billingham Division) Report BR-296, 1944 
^ DusHMAN, S. Scientific Foundations of Vacuum Techniques, p 100. Wiley, New 

York, 1949 
" Clausing, P. Ann. Phys. 12 (1932) 961 
'' Guthrie, A. and Wakerling, R. K. Vacuum Equipment and Techniques, p. 36. 

McGraw-Hill, New York and London, 1949 
^ PiRANi, M. and Yarwood, H. Principles of Vacuum Engineering, pp 11-12. 

Chapman and Hall, London, 1961 
' Ward, L. and Bunn, J. P. Some Aspects of the Design of High Vacuum Systems. 

Engineering Materials and Design Association, London, 1965 
1° Davis, D. H. J. appl. Phys. 31 (1960) 1169 
" Levenson, L. L., Milleron, N. and Davis, D. H. Vacuum Symp. Trans. 

(1961) 372 
12 Smith, C. G. and Lewin, G. J 2th Natn. Symp. Am. vacuum Soc. (1965) Paper 

10.1 



36 



THE PRODUCTION OF VACUUM 



3.1 Mechanical Oil Sealed Rotary Pumps 

The starting point in the production of a high vacuum in many 
systems is a mechanical pimip. Such pumps are capable of produc- 
ing pressures of the order of 10"^ torr, and are carefully designed, 
finely engineered pieces of equipment which should be handled and 
maintained with great care. The basic principle of operation of these 
pumps is outlined in the following discussion. 

The system to be evacuated is opened to the pump chamber whose 
volume is then increased by mechanical movement of a piston. Thus 
the volimie of the system gas is increased, causing a pressure reduction 
according to Boyle's Law. The gas in the pump is then isolated from 
the system and compressed to atmospheric pressure when it can be 
discharged to the atmosphere. This cycle of operations is repeated 
bringing about a pressure reduction at each cycle. In theory, the 
low pressure limit of such a device is determined only by the fact that 
the gas is compressed into a small but finite 'dead space' volume. 
When the system pressure becomes so low that, at maximum compres- 
sion, the gas pressure is still less than that of the atmosphere it cannot 
be discharged from the pump. Subsequent pumping action re- 
expands and recompresses the same gas without drawing any more 
from the system. The ratio of the exhaust pressure to the inlet pres- 
sure is termed the pump compression ratio. Thus, to produce 
pressures of the order of 10"^ torr, pumps having compression ratios 
of the order of 10® are required. 

3.1.1 Types of Oil Sealed Pump 

There are several variations of this type of pump currently in use, 
which differ only in the details of application of the basic principles 
given above. One of these is described fully and the remainder 
are discussed briefly to point out their particular design features. 

3.1.1.1 The Rotating Vane Type — The essential parts of this type of 
pump are illustrated by the exploded view of Figure 3.1 and by a 

37 



' 



THE PRODUCTION OF VACUUM 





S.e 








<D a 


01 


q^ 














a 


"D O 


JZ 


c x: 


in 


UJ 01 




f 



t 



<3 



>«, 



3 
•S 

(J5 



.60 



38 



MECHANICAL OIL SEALED ROTARY PUMPS 

vertical section in Figure 3.2. The stator is a hollow steel cylinder the 
ends of which are closed by suitable plates. It is pierced by the inlet 
and exhaust ports which are positioned respectively a few degrees 
on either side of the vertical. The inlet port is connected directly to 
the system by suitable tubulation while the exhaust port is covered by 
a valve. In older pumps the exhaust valve is a metal plate which can 
move vertically between the outside of the stator and an arrester 
plate. The more modern type consists of a rectangular sheet of 
synthetic rubber (usually Neoprene) which is constrained to hinge 
between the stator and a metal backing plate. 



Exhaust 
outlet 



Oil splash 
baffles 



Inlet tube 



Flap valve 
backing plate 

Exhaust 
flap valve 




Inlet port 
Top seal 



Figure 3.2. Vertical section through rotating vane pump 

The rotor consists of a steel cylinder mounted on a driving shaft 
which passes through one of the end plates. Its axis of rotation is 
parallel to, but displaced from, the axis of the stator such that it makes 
contact (to within about 0-001 in.) with the top surface of the stator, 
the line of contact lying between the two ports. This line of contact is 
known as the top seal. A diametrical slot is cut through the length 
of the rotor and carries the vanes. These are rectangular steel plates 
which make a sUding fit in the rotor slot and are held apart by springs 
which ensure that the rounded ends of the vanes always make contact 
with the stator wall. In the case of small pumps the whole of the 

39 



THE PRODUCTION OF VACUUM 



stator/rotor assembly is submerged in a suitable oil. If necessary, larger 
quantities of oil may be fed from a reservoir to the appropriate regions. 
The action of the pump can be followed by reference to Figure 3.3. 
As vane A passes the inlet port, as in Figure 3.3 [a), the system is con- 
nected to a space (that between the stator, the rotor, vane A and the 





(a) 



Volume (at system pressure) 
swept per half revolution 

(6) 








(0 



(d) 



Figure 3.3. Rotating vane pump illustrating mode of action: [a) 
induction, {b) isolation, (c) beginning of compression, [d) exhaust 

top seal) whose volume increases as the vane sweeps round, thus 
producing a pressure reduction in the system. This continues until 
vane B passes the inlet port indicated in Figure 3.3(b), when the in- 
crease of volume of the system gas is that between the two vanes. 
The induction stage is now complete and the gas is isolated from both 
the inlet and the outlet ports. Further rotation sweeps the isolated 
gas around the chamber until vane A passes the top seal as shown in 
Figure 3.3(c). The gas is now held between vane B and the top seal 

40 



MECHANICAL OIL SEALED ROTARY PUMPS 

and its volimie is reduced as this vane approaches the top seal. 
Eventually the pressure within this space becomes sufficient (about 
850 torr) to open the exhaust valve, and the gas is forced out of the 
pump chamber. During the isolation and exhaust stages further gas 
has been drawn into the pump behind vane B and is just about to 
enter the compression stage, as in Figure 3.3(d). 

In one revolution of the rotor a volume of gas equal to twice that 
indicated in Figure 3.3(b) is displaced by the pump. Thus, the 
product of this volume and the number of revolutions of the rotor per 
unit time is the volume rate at which gas is swept round the pump. 
This rate is termed the pump displacement 5o. 

It can be seen that the contacts of the vanes and rotor with the stator 
form three separate chambers each in general containing gas at dif- 
ferent pressures. These contacts must therefore make vacuum-tight 
seals, especially for the top seal which must support the differential 
between system and atmospheric pressures. For this reason the inner 
surfaces of the stator and all rotor and vane surfaces are very carefully 
machined to reduce potential leakage paths to a minimum. Hence, 
great care must be taken to ensure that no abrasive material or gas 
which is likely to corrode the metal surfaces enters the pump chambers. 
Care is also required when storing and handling the pump components 
during maintenance. 

In order to complete the contact seals and to lubricate them, small 
and carefully controlled quantities of oil are allowed to enter the pump 
chamber where it is guided by channels and grooves to the contacting 
surfaces. 

In addition to lubrication and sealing, the oil also performs a third 
function, namely that of helping to accomplish the very high compres- 
sion ratios required at low inlet pressures. It is argued on p 37 that 
at low inlet pressures the exhaust valve will not open because of the 
dead volume between the valve and the piston (i.e. the vane). Now, 
if there is sufficient oil in the compression/exhaust chamber the dead 
volume becomes filled with oil at, or just before, the gas (whose 
volume is less than that of the dead space) reaches exhaust pressure. 
Further movement of the piston then forces the oil and the bubble of 
gas through the exhaust valve. If there is insufficient oil the required 
compression ratio is not obtained, but if there is an excess the oil is 
occupying pump volume which could otherwise be occupied by gas. 
It can be seen further that the lower the inlet pressure the greater is the 
quantity of oil required, and thus the rate at which oil enters should 
be governed by the inlet pressure. This is achieved by boring a hole 
through the stator wall into the inlet chamber. The rate of flow of 
oil through this hole is dependent upon the hole length and radius, 

41 



THE PRODUCTION OF VACUUM 

the oil viscosity and, in particular, the pressure difference across the 
hole {i.e. atmospheric less inlet pressure). Consequently, oil enters 
the pump chamber at a relatively high rate when the inlet pressure is 
low and at a low rate when the inlet pressure is high. 

A result of having the oil hole into the inlet chamber is that if the 
pump is switched off and the system left at low pressure oil is forced 
into the inlet chamber and eventually into the system. To avoid this 
the pump must be opened to atmospheric pressure as soon as it is 
switched off. This may be accomplished by a manually operated air 
admittance valve immediately above the pump, or by an electro- 
magnetically operated valve which opens whenever the power supply 
to the pump is cut. A recent development incorporates a small oil 
pump to inject oil into the pump chambers. When the vacuum 
pump is not operating the oil pump acts as a valve preventing the 
flow of oil into the system. 

3.1.1.2 The Sliding Vane Type — A vertical section through this type of 
pump is illustrated in Figure 3.4. The pump has a single vane which 



Inlet por 




Oil splash baffle 
Exhaust 

Vane 

Stator 

-Rotor 



Figure 3. 4. Vertical section through a sliding vane rotary pump 

slides in a slot cut in the stator between the inlet and exhaust ports. 
The movement of the vane is controlled via a rocker bearing in the 
outer sleeve of the rotor, this sleeve being a sliding fit on the rotor 

42 






1 



MECHANICAL OIL SEALED ROTARY PUMPS 

proper. The latter rotates eccentrically about the central axis of the 
stator such that the sleeve moves around in contact with the stator. 
The whole assembly is submerged in oil, controlled quantities of which 
are allowed to enter the pump chamber to complete the vacuum 
seals and to provide lubrication around the vane and at the rotor/ 
statoi' contact. 







Figure 3.5. Sliding vane rotary pump illustrating mode of action : 
(a) induction, (6) isolation, (c) compression, {d) exhaust 

The pumping cycle is illustrated in Figure 3.5. It can be seen that 
the isolation stage occurs instantaneously as the rotor passes the vane 
slot, and that the volume of gas swept around the pump at each 
revolution is that between the stator and rotor at this instant. 

3.1.1.3 The Kinney Type — ^The physical principles of operation of this 
pirnip are basically the same as those of the sUding vane type, al- 
though mechanicallv there are several differences. The pump is 
designed to handle k%e quantities of gas; hence its working volume is 

^ 43 



THE PRODUCTION OF VACUUM 

quite large and as a result it is possible to combine the functions of the 
inlet port and the sliding vane. 

The essential features of the pump and its action are illustrated in 
Figure 3.6. The sliding vane is replaced by a hollow tube which is 
rigidly attached to the outer sleeve of the rotor. The tube rolls and 
slides in a rocker bearing and a hole cut in the tube allows gas to be 
drawn into the inlet side of the pump. Because of the large quantity 
of gas to be exhausted the exhaust ports are of relatively large diameter. 
The simple rubber flap valve fitted to the sliding and rotating vane 



Gas inlet 



Cooling 

water 

outlet 




Gas i 



Exhaust 



Cooling — ' 
water inlet 




Exhaust 



Cooling 

water 

inlet 



-Cooling 
water outlet 



(a) (b) 

Figure 3.6. The construction and action of a Kinney pump : (a) induction, (b) exhaust 



type pumps is not sufficiently strong to provide adequate closure of 
these large ports, and so spring loaded poppet type valves are used 
instead. The heat of compression of the gas in these large pumps 
can be considerable and so the stator is usually provided with a 
jacket through which cooling water may be circulated. 

3.1.2 Ultimate Pressure 

When in good working order and used on a clean, leak-free system, 
the pimips described above will produce an ultimate pressure of the 
order of 5 x 10"^ torr, regardless of their displacement. This limit 
is imposed almost entirely by the failure of the oil seal between the 
inlet and exhaust chambers, with the result that during compression 

44 



MECHANICAL OIL SEALED ROTARY PUMPS 

gas flows across the seal and re-expands in the inlet chamber and 
hence into the system, leading to no further pressure reduction. This 
effect can be considerably reduced by connecting two pumps in 
series so that the pump which ' sees ' the system has a comparatively 
low exhaust pressure (2-3 torr) and thus a small pressure difference 
between inlet and exhaust chambers. The two pumps are usually 
mounted as one unit with a common driving shaft, to ensure that the 
respective pumping cycles remain in correct phase. Such a device is 
known as a two-stage pump and can achieve ultimate pressures 
approaching 10 ~* torr. 

The foregoing discussion assimies that the gas being pumped is 
clean. However, there are several ways in which a gas may be 
considered dirty from a vacuum point of view. For example, if the 
gas carries fine abrasive material or a corrosive component the stator 
and rotor surfaces may be damaged causing leakage paths between 
inlet and exhaust chambers, giving a gradual and permanent deterior- 
ation in ultimate pressure. Further, if the gas has a component 
with which the oil reacts, the oil will become less effective as both a 
sealing medium and a lubricant, the latter defect leading to permanent 
damage. 

Condensable vapour components of the gas are another, and very 
common, cause of poor pump performance. The vapour is normally 
present in an unsaturated state and is satisfactorily pumped until the 
compression required to open the exhaust valve is sufficient to raise 
the vapour pressure to its saturation value. At this stage the vapour 
begins to condense to a liquid and, on exhaust from the pump 
chamber, remains mixed with the oil. This can have two effects : it 
may emulsify or react chemically with the oil, impairing its lubrica- 
tion and sealing properties; or it may circulate with the oil to re- 
evaporate on the inlet side of the pump. Unless the vapour is 
removed the efl^ect is cumulative and eventually the lowest pressure 
obtainable is the s.v.p. of the liquid. Water vapour is the greatest 
cause of trouble in this respect, the vapour appearing as a result of 
vessel outgassing rather than as that originally present in the vessel 
atmosphere. 

As an example, consider a system in which the dry air pressure is 
1 00 torr with 1 torr water vapour pressure and a mean pump tempera- 
ture of 30°C. The compression ratio to exhaust this to the atmos- 
phere is about 850/101 = 8-5. Thus, the vapour pressure on exhaust 
is about 8-5 torr which is less than the s.v.p. of water vapour at 30°C. 
However, with continued pumping the dry air pressure may fall to 
10 torr, but because of outgassing the water vapour pressure may have 
fallen to only 0-9 torr. A similar calculation leads to a vapour 

45 



THE PRODUCTION OF VACUUM 

pressure on exhaust of about 76 torr, which is much greater than the 
s.v.p. of water vapour, and so condensation will occur. 

3.1.3 Removal of Vapours 

There are several ways in which the problem presented by condens- 
able vapours may be solved. They may be removed by physical or 
chemical means before reaching the pump, or by the pump itself if 
suitably modified. 

3.1.3.1 Chemical Removal — A trap is provided, containing a material 
with which the vapour will react without forming undesirable by- 
products. The disadvantages of this method are that (i) generally a 
separate material is required for each vapour, (m) the material must 
be replaced as soon as it has all reacted, and (m) the original material 
cannot be easily reclaimed. The commonest example of this method 
is the use of phosphorus pentoxide to remove water vapour by the 
reaction 



P2O5 + 3H2O 



2H3PO, 



Water is removed to the extent that the vapour pressure associated 
with the reaction products is of the order of 2 x 10~^ torr. 

3.1.3.2 Physical Removal — ^A trap is provided in which the vapour is 
either sorbed on the surface of a solid or is cooled to a temperature 
(usually to its solid phase) at which its s.v.p. is extremely low. The 
sorption method again suffers from the disadvantage that the sorbent 
must be replaced periodically since it can become saturated with 
vapour, but it can usually be reclaimed fairly easily. The use of 
silica gel to trap water vapour is a common example of this method. 
The cooling method has the advantage that it is reasonably effective 
for all vapours, the commonest coolants being solid CO2 (in acetone or 
methylated spirit) or liquid nitrogen. Liquid nitrogen is preferable 
since at its boiling point ( — 196°C) the s.v.p. of water is about 10"^^ 
torr, whereas at the temperature of solid CO2 ( — 80°C) the s.v.p. of 
water is 6 x 10"* torr. 

3.1.3.3 Gas Ballasting — ^This technique of vapour removal involves a 
modification to the pump, and offers considerable advantages over 
the methods described above. It requires the minimum of attention, 
involves no extra running cost and only a small increase in capital 
cost, and is effective for all vapours except those which dissolve in the 
oil. Its disadvantages are that there is an upper limit to the inlet 

46 



MECHANICAL OIL SEALED ROTARY PUMPS 

vapour pressure and to the rate at which the vapour can be handled, 
and that the ultimate pressure obtainable is considerably increased. 

Table 3.1. Summary of Methods oj Vapour Removal 



Method 


Vapours for which 
effective 


Advantages 


Disadvantages 


Chemical 


Selected 


Permanent 
removal of 
vapour 


Periodic main- 
tenance required; 
cannot be 
recovered 


Sorption 


All, to some extent, 
but most effective 
for selected 
vapours 


Can be recovered 


Periodic main- 
tenance 


Refrigerant 


All 




Periodic main- 
tenance 


Gas ballast 


All 


Permanent removal 
of vapour; no 
maintenance 


Upper limit to 
vapour pressure 
and rate of flow ; 
high ultimate 
pressure 



In the gas ballast technique, a controlled amount of atmospheric 
air is admitted to the exhaust chamber of the pump. The pressure 
in this chamber is thus raised sufficiently to open the exhaust valve 
without the necessity for extreme compression of the system gas and 
its vapour content. Consequently, the pressure of the vapour is 
not increased to its saturation value and the vapour is swept out of the 
pump with the permanent gas. 

A typical method of admitting ballast air is illustrated in Figure 3.7. 
The ball valve only opens and admits ballast air if atmospheric 
pressure is greater than the sum of the pressure equivalent of the 
spring tension, and the pressure of system gas in the exhaust chamber. 
The spring tension, and hence the amount of ballast air entering the 
pump, can be varied by manual adjustment of the valve seat. 

To illustrate the action, suppose the valve is set such that the spring 
tension is equivalent to 100 torr pressure; atmospheric air then 
enters the exhaust chamber until the total pressure is 660 torr (assum- 
ing atmospheric pressure to be 760 torr). It is then only necessary 
for the pump to apply a compression ratio of 850/660 = 1-3 in order to 
open the exhaust valve. Applying this to the example of Section 
3. 1 .2 and assuming the ballast air to be completely dry, the pressure 

47 



THE PRODUCTION OF VACUUM 

of water vapour when the exhaust valve opens is 1-3 x 0-9 = 1-16 torr 
and hence the vapour is not condensed. 

Gas ballast is also effective in removing condensed vapours from oil 
which has been allowed to become contaminated by operating the 
pump when it is isolated from the system. Power and Kenna<" 
claim that gas ballast is also partially successful in removing gases and 
vapours which may have dissolved in the oil. 

The ultimate pressure obtainable during ballasting is increased 
because there is a high pressure difference across the top seal for a 
much longer fraction of the compression/exhaust stage, resulting in 
greater time for leakage. In some pumps this is overcome to some 



To atmosphere 




Oil level 



Channel to 
exhaust side 
of pump 



Figure 3. 7. Typical gas ballast valve 



extent by providing means for admitting extra oil to the pump and 
hence improving this seal. Another possibility is to operate the 
pvmip on a 'dirty' system without gas ballast and so obtain reasonable 
ultimate pressures until the oil is so contaminated that the pump is no 
longer effective. The pump is then isolated from the system and put 
on full gas ballast in order that the oil may be decontaminated. If 
necessary, a second pump is used to take over the normal pumping 
duties until the first is again usable. 

A theoretical and experimental investigation of the operation of gas 
ballast pumps was carried out by Power and Kenna'^*. Their work 
shows that the maximum mass flow rate of vapour i?„ is given by 



R,: = 



T\P,S+{Po-pu)Sa 



iPe- 

48 



■ps) 



(3.1) 



MECHANICAL OIL SEALED ROTARY PUMPS 



and that the maximum pressure of vapour /)„ at the pump inlet is 
Ps Tlp^+{Po-pn)S, S,p^ 



" s tX 



Ps J 



(3.2) 



iPe-ps) 

where po is the total pressure of the ballast air, 

p^ is the partial pressure of vapour in the ballast air, 

Sa is the speed at which ballast air enters, 

pg is the partial pressure of air at the pump inlet, 

Ps is the s.v.p. of the vapour^ 

Ps is the density of the vapour atp^, 

S is the pump speed at the inlet, 

pg is the pressure at which the exhaust valve opens, 

T (°K) is the pump temperature, 

Tq (°K) is the ambient temperature. 
The experimental work shows that the theoretical values are of the 
correct order but are somewhat conservative. 

3.1.4 Oil Sealed Pump Characteristics 

The essential vacuum characteristics of these pumps are sum- 
marized by curves showing the variation of speed (as defined in 



500 



AOO 



c 

E 300 



^ 200 



100 



<Sn = 500 Lmin"^ 



S=Sq(1-Pu/p) 
tor So = 3001.nnin 



10"' 




Sn=300l.min 



5n= 160 l.min 



10^ 



10' 



Figure 3.1 



Pressure, torr 
Typical rotary pump characteristics 



Section 1 .4) at the pump inlet with inlet pressure. Since the pressure 
axis covers a large range it is usually plotted on a logarithmic scale. 
It is also convenient to plot the speed on the same scale since then 
diagonals represent lines of equal throughput. Figure 3.8 shows some 



4 I.H.V.T. 



49 



THE PRODUCTION OF VACUUM 

typical pump characteristics. It can be seen that (i) all single-stage 
pumps have much the same ultimate pressure regardless of their 
displacement, (ii) over quite a wide range of pressure the speed is 
essentially constant and equal to the pump displacement, and (m) 
throughput falls very rapidly with falling pressure especially when the 
speed is also falling. 

An approximate equation relating the speed and pressure can be 

derived as 



S = S,{l-pJp) 



(3.3) 



where S is the speed at pressure/), ^o is the displacement, and /)„ is the 
ultimate pressure. The difference between this relationship and that 
observed experimentally is shown in Figure 3.8 for the case of a 300 
1. min-^ pump. The use and interpretation of these characteristics 
is important in the design of vacuum systems and is dealt with in 
Chapter 6. 

Pumps having displacements in the following ranges are com- 
mercially available : 

rotating and sliding vane types, 20-500 1. min"^, 

Kinney type, 300-20,000 1. min'^ 

3.2 The Roots Pump 

This is a mechanical pump based upon a design originally intended as 
a blower. It consists of a hollow, cylindrical stator with diametrically 
opposed inlet and exhaust ports. Mounted within the stator are two 
double lobed rotors, or impellers, which rotate at the same speeds but 
in opposite directions. Since these pumps can handle very large 
quantities of air, water-cooled tubes are often placed in the exhaust 
region to remove the heat of compression. The impellers are also 
often of hollow construction with means for circulating cooling water 
via the drive shafts. Figure 3.9 illustrates the construction and action 
of the pump. 

Air from the system fills the space shaded in Figure 3.9(a) between 
the impellers and the stator. As the impellers rotate the upper one 
traps a portion of this gas between itself and the stator (giving some 
degree of compression) and sweeps it round towards the outlet as in 
Figure 3.9{b). When the leading lobe of the upper impeller passes the 
exhaust port, air from the discharge region (at a relatively high pres- 
sure) mixes with the gas which has been carried from the inlet, 
as shown in Figure 3.9(c). Further rotation of the upper impeller 
compresses the gas against the upward moving lobe of the lower 
impeller, so forcing the gas out through the exhaust port. 

50 



the roots pump 



The impellers do not touch the rotor or each other; in fact, there is a 
clearance of about 0-01 in. in both cases. Thus, there is a back flow 



Exhaust 




Inlet Exhaust 




Exhaust 




Inlet Exhaust 




'(c) \-. ^ id) 

Figure 3.9. Roots pump, illustrating its action 



20,000 



15,000 



, 10,000 



5,000 




10 






lO" 



10' 



lO'^ 



10""^ ' 10"' 10° 
Pressure, torr 
Figure 3.10. Typical Roots pump characteristics 



of gas from the discharge region to the inlet region and therefore the 
efficiency of compression is much lower than in the case of oil sealed 
pumps. However, the absence of rubbing contacts means that higher 

51 



THE PRODUCTION OF VACUUM 

speeds of rotation are possible, leading to much higher pumping 
speeds. Since the conductance of the gaps decreases as the mean 
pressure in the pump falls, the pump efficiency would be expected to 
increase with falling pressure, and this is in fact the case. Both 
experimentally and theoretically'^', maximum efficiency occurs 
when the pump is operated at a compression ratio of 10 at a pressure 
of the order of 5 x lO'^ torr, and thus the pump must be provided 
with a suitable backing faciUty. Roots pumps with maximum speeds 
up to 2xl0^1.min-i ^re available. A typical speed /pressure 
characteristic is illustrated in Figure 3.10. 

3.3 Vapour Pumps 

It was seen that the ultimate pressure obtainable with mechanical 
pumps is of the order of 1 " ^ torr. In order to obtain lower pressures 



Umbrella 



Chimney. 

To backing 
pump 



Vapour 
molecules 



Boiler 




Cooling 
water coils 



//^//J^A^^^^AMAA^w^AA/^ — 



Figure 3.11. Single-stage vapour pump 

it is necessary to use non-mechanical pumps and one of these is the 
vapour diffiision pump. The pumping action is achieved by impart- 
ing to the gas molecules from the system a momentum in a direction. 

52 






VAPOUR PUMPS 

towards the exhaust, by meansof a high speed stream of vapour mole- 
cules. The principle of operation and construction is illustrated in 
Figure 3.11. 

The pump consists essentially of a cylindrical body at the base of 
which is a boiler containing a suitable fluid. Provision is made for 
cooling the upper part of the body either by a water jacket or by fins 
in an air stream produced by a fan. A chimney projects into the 
boiler and mounted above the chimney is an umbrella shaped 
deflector or jet. Vapour molecules from the boiling fluid travel up 
the chimney and are deflected by the umbrella into a downward 
moving stream travelling at high speed. Gas molecules which enter 
the vapour stream receive considerable downward momentum as a 
result of collisions and are swept towards the exhaust. The vapour 
eventually collides with the cool wall where it condenses to liquid and 
flows back to the boiler. 

3.3.1 Multistage Pumps and Jet Design 

The single-jet pump described above does not function very 
efficiently in practice. Although the theory of the vapour pump is 
not fully understood the following discussion indicates the reasons 
for the inefficiency and the basis of practical pump design. 

In order to achieve a high pumping speed the pump must 'see' a 
large area of the system so that gas molecules enter the vapour stream 
at a high rate. Hence the jet 'admittance area' (nominally the area 
of the annulus between the jet and the pxmip wall) must be large. 
However a large admittance area implies a large escape area for 
' back-diffusion ' of gas through the vapour stream and into the system. 
Back-diffusion is at its maximum near the pump walls, where the 
vapour stream is of low momentum and of low density due to scatter- 
ing at intermolecular collisions. The effect is more pronounced at 
high intake pressures [i.e. high gas molecule densities) since the vapour 
molecules then make a large number of collisions within a very short 
distance of the jet. Since the pumping speed is the net rate of removal 
of gas from the system, a high back-diffusion rate impUes a low pump- 
ing speed and olaviously when the back-diffusion rate equals the for- 
ward-diflfusion rate the pumping speed is zero. 

From the foregoing argument it is apparent that for efficient 
operation of the pump two conditions must be fulfilled: 

{a) The system pressure must be initially reduced below a certain 
value which, in most practical cases, is of the order of 10"-^ torr. 

{b) The pressure below the jet must be kept reasonably low to 
reduce the probabiUty of back-diffusion. 

To achieve these conditions vapour pumps are constructed with 

53 



THE PRODUCTION OF VACUUM 

several jet stages in series, one acting as a backing pump to another. 
Such a pump is illustrated schematically in Figure 3.12. It can be 
seen that the jets project nearer to the pump wall the lower the jet 
considered. The main function of the top jet is to give a large pump- 
ing speed and thus this jet has a large admittance area. On the other 
hand the lower jets have smaller admittance areas, and hence smaller 
escape areas, and the vapour streams are generally denser. Conse- 



Speed 



Pressure 



100 I. sec"' 5 X 10 ^ torr 



^y^ 



10 1. sec"' 5 X 10 '^ torr 



5 I. sec 



10"^ torr 



To backing^ 
pump 



f 



I L-i n n 



001 l. sec"' 5x10 ' torr 



Figure 3.12. Three-stage self-fractionating pump 

quently the speeds of the lower jets become successively smaller whilst 
the pressure differences which they can support become larger. The 
speed and pressure data in Figure 3.12 illustrate these points. It 
should be noticed that the throughput is necessarily constant through- 
out the pump. 

3.3.2 The Pump Fluid 

Ideally this should be of high molecular weight so that each mole- 
cule carries considerable momentum and can therefore make effective 
collisions with several gas molecules before all of its momentum is lost. 

54 



VAPOUR PUMPS 

It should also have a very low vapour pressure at room temperature 
since molecules enter the system both by evaporation from the con- 
densed liquid on the pump walls and by 'back-streaming' {i.e. the 
migration into the system from the vapour stream due to inter- 
molecular collisions). It should be resistant to chemical attack at 
boiler temperature, and under prolonged heating should resist 
'cracking' into lighter (and generally higher vapour pressure) 
fractions. These conditions impose severe limitations upon the fluids 
which may be used. Those in common use, and some of their 
properties, are given in Table 3.2. 

Table 3.2. Properties of Some Vapour Pump Fluids 



Fluid 


Molecular 
weight 


Vapour 

pressure at 

25°C 

(torr) 


Boiling 

point at 

10-^ torr 

(°C) 


Resistance 

to 
oxidation 


Resistance 

%■ 
cracking 


Mercury 


200 


10-3 


47 


Good 


Perfect 


Hydrocarbon oils: 
Apiezon A* 
Apiezon B* 
Apiezon C* 


410 
470 
570 


2x10-5 

4x 10-'' 

10-8 


110 
110 
125 


Poor 


Poor 


Silicone oils : 
DC 702t 
DC 703t 


530 
570 


io-« 

10-' 


160 
200 


Excellent 


Excellent 



♦ Shell Chemical Co. Ltd. f Dow Corning Corp. 

It can be seen that of these fluids the siUcone oils offer the greatest 
advantages for normal use although for certain applications they are 
unsuitable. Mercury has a high vapour pressure necessitating the use 
of a cold trap. It is common practice to use a cooled baffle immediately 
above the ptmip to condense mercury to liquid so that it may return 
to the pump. A liquid nitrogen trap is used above the baffle to 
reduce the vapour pressure to a very low value (~ 10"^'' torr). 

The poor resistance of the hydrocarbon oils to oxidation and 
cracking is a serious disadvantage and as a result they have been 
largely superseded by the silicone oils. A discussion of the relative 
merits of mercury and oils, from the point of view of their suitability 
for particular applications is given in Chapter 6. 

3.3.3 Self-fractionation of the Pump Fluid 

Further reference to Figure 5. 72 shows that the bases of the chimneys 
feeding the various jets divide the fluid in the boiler into a series of 

55 



THE PRODUCTION OF VACUUM 

interconnected annular zones. Fluid returned to the boiler from all 
the jets enters the outermost zone and to reach the central one it must 
pass through the other zones. The path through the various zones is 
made tortuous, so that the fluid spends a relatively long time in each 
and the high vapour pressure components are boiled off to chimneys 
feeding the lower jets. Thus fluid which reaches the central zone has 
the lowest vapour pressure. 

3.3.4 Cooling 

This should be arranged so that the coolest region is where the 
vapour stream strikes the pump wall. The rate of cooling is also 
fairly critical since if it is too low the vapour is not entirely condensed 
and thus the back-streaming effect is increased. On the other hand, 
if the cooling rate is too high the vapour is not only condensed but also 
considerably cooled. This results in a slow flow back to the boiler, 
and necessitates a high boiler power to re-evaporate it. 

3.3.5 Backing and Roughing Requirements 

In order to provide the conditions under which a vapour pump can 
operate, gas must be removed from the pump exhaust at a rate such 
that the exhaust pressure does not rise above a certain value (the 
critical backing pressure p,,) . If this is not achieved, excessive back- 
diffusion occurs at each stage of the pump thus reducing, or even 
stopping, the pumping effect. For oil vapour pumps, critical backing 
pressures are of the order of 10"^ torr but for mercury vapour pumps 
they can be several torr. 

It is also necessary to provide an initial reduction in system pressure 
before the pump can begin to operate. This initial pressure reduction 
is known as roughing and must obviously be carried out to pressures at 
least as low as the critical backing pressure. 

3.3.6 Speed Characteristics and Ultimate Pressure 

Provided roughing and backing facilities are adequate, the vapour 
pumping action reduces the system pressure, thus decreasing the 
density of gas molecules entering the vapour stream. This in turn 
reduces back-diffusion and hence the pumping speed rises with 
decreasing pressure. The speed continues to increase until the pres- 
sure is such that the rate of back-diffusion at the top jet is controlled 
not by the inlet gas density but by the rate at which gas is removed 
from below the jet. At this, and lower system pressures, the speed 
remains constant at a maximum value QiS„. Thus for a given top jet 
admittance area, the provision of extra stages would be expected to 

56 



VAPOUR PUMPS 

give higher values of S^ or alternatively a higher critical backing 
pressure. 

The ultimate pressure obtainable is theoretically determined by the 
vapour pressure of the pump fluid. With additional equipment 
(refrigerated baffles, cold traps, etc.) much lower pressures can be 
obtained. In practice however the ultimate pressure is governed by 
the characteristics of the system, in that the pumping speed in the 
system must necessarily become zero when the rate of gas generation 
(by leaks, outgassing, etc.) is equal to the maximum rate at which the 



in 

















1,000 








"^ 


\ 












\ 


V 


100 
10 


D 




C 




r> 


\ 




1 

1 
1 

1 


y 






\ 




1 

1 








\ 


1 




1 











10" 



10" 



10"^ 10"'' 



10"' 



10'' 



10" 



Pressure, torr 
Figure 3.13. Typical vapour pump characteristics 



pump can handle gas. This is illustrated by the typical speed charac- 
teristics shown in Figure 3. 13. The curve ABCD shows the theoretical 
characteristic with the speed, at pressures below p^, remaining 
constant at S^ = 1001. sec" ^. The curve ABCE shows a typical 
practical characteristic in which the pumping speed becomes zero at 
/>„ = 5x 10"^ torr when the' throughput is 5 x 10"* torr 1. sec" ^. 
The characteristic of a pump for which S^ = 10,0001. sec"^ is also 
shown. 

The section AB of the characteristic is reasonably linear on a log 
log plot and hence may be represented by the empirical equation 



•^S 



iP > Pm) 



(3.4) 



where z is the slope of the line AB taking values between 0-8 and 1 . 
The section BCE can be represented by an equation similar to that 

57 



THE PRODUCTION OF VACUUM 

governing the speed variation of a rotary pump, thus 



5=^l-^j {P<Pm) 



(3.5) 






Vapour pumps having values oiS^ in the range 10-45,000 1. sec" ^ are 
commercially available. The physical sizes of the slowest and fastest 
pumps are roughly 1 in. diameter by 12 in. high, and 36 in. diameter 
by 72 in. high. 

3.4 Vapour Booster Pumps 

By comparing the vapour pump characteristic oi Figure 3.13 with that 
of typical rotary pumps in Figure 3.8, it can be seen that in the pressure 
range 10" -"^-lO"^ torr both rotary and vapour pumps are working at 
low speeds, thus limiting the throughput which can be handled 



High speed 
jet 



Cooling water 
coils 



Cooled 
baffle- 
plates 




Boiler- 



rTW^AAAW/pl 



Figure 3.14. Two-stage vapour booster pump 

58 



VAPOUR BOOSTER PUMPS 

economically. To overcome this, modified vapour pumps known as 
booster pumps were designed to have high speeds in this region but 
without a particularly low ultimate pressure; they also have a high 
critical backing pressure. To illustrate the action of such a pump a 
two-stage version is illustrated in Figure 3.14. 

Since a low ultimate pressure is not required an oil of relatively high 
vapour pressure (~ 10"* torr) is used, enabling particularly dense 
vapour streams to be obtained without the necessity of excessive 
boiler power. The top jet of the pump projects somewhat nearer the 
pump walls than in the normal vapour pump thus giving increased 



10,000 



1.000 



-D 
<ll 
<D 
Q. 

U1 




Pressure, torr 
Figure 3.15. Typical vapour booster pump characteristics 

resistance to back-diffusion. This, together with the very dense 
vapour stream, means that even at high system pressures the back- 
diffusion rate is small compared with the forward-difTusion rate, and 
hence the pump will operate at high pressures. 

The gas removed from the system is further compressed, by virtue 
of the convergent pump walls, before it enters the second (or ejector) 
jet. This jet is a divergent nozzle which ejects the gas directly into 
the backing line and provides a very dense and high speed vapour 
stream with small admittance and escape areas. This stage is there- 
fore of low speed but is capable of supporting a large pressure dif- 
ference. Further gas compression is again obtained by means of the 
convergent walls, with the net result that a low backing pressure is not 
required. To ensure the complete condensation and return of pump 

59 



THE PRODUCTION OF VACUUM 

fluid from the ejector stage the exhaust tube is vertical and contains 
baffle plates thermally connected to the cooling water tube. 

Higher pumping speeds may be obtained by adding further stages 
between the top jet and the ejector. Ejector stages are sometimes 
added to the normal high speed vapour pump discussed in Section 
3.3 to give them a higher critical backing pressure. 

Figure 3.15 shows the characteristics of three typical vapour booster 
pumps. Critical backing pressures are about 5 torr. The charac- 
teristics are of similar shape to those of normal vapour pumps and may 
be represented by similar equations (see eqns 3.4 and 3.5). 

Values ofS^ for commercially available vapour booster pimips range 
from40 to 25,0001. sec- 1. 

3.5 Sorption Pumps 

If a previously degassed material is exposed to the gas in a vessel, gas 
will be readily sorbed by it according to the theory outlined in Section 
1.11.1. If the area of the exposed surface is large and its temperature 
is low the quantity of gas sorbed will be considerable and will cause a 
significant reduction of the pressure in the vessel. 

The sorption pump consists essentially of a cylindrical vessel, 
usually made of stainless steel and containing an activated sorbent 
material, which is surrounded by a Dewar vessel of liquid nitrogen. 
On opening the pump to the system, gas is sorbed until the sorbent is 
saturated and the pressure has fallen to an equilibrium value. This 
pressure is determined by the sorptive capacity {i.e. the total quantity 
of gas which can be sorbed) of the pump relative to the quantity of 
gas initially in the system. The sorptive capacity of a given material 
varies from gas to gas, and hence the ultimate pressure is different for 
different gases. Sorptive capacities for the inert gases tend to be low 
since these gases can only be physically adsorbed; thus when they 
undergo sorption pumping they are considerably enriched. 

The theoretical speed of such a pump is proportional to the exposed 
area of sorbent and can be high, but in practice it is usually limited by 
the impedance presented by the pump body and the pipe line connect- 
ing the pump to the system. 

Sorption pumps are generally used to pump from atmospheric 
pressure, and ultimate pressures of the order of 10""^ torr are achieved 
provided the sorptive capacity is correctly matched to the volume of 
the system. They may also be used to obtain very low pressures if the 
system is previously rough pumped. For example, Read*^' describes 
the use of a sorption pump which pumped from 5xl0-''tolxl0-^ 
torr in 50 min. 

60 



SORPTION PUMPS 

The materials used as sorbents have a fine pore structure and thus 
have extremely large effective areas for a given volume. Those in 
common use are charcoal, alumina and molecular sieves. Molecular 
sieve materials are usually artificial zeolites (metal aluminosilicates) 
and are described by Hersch<^> and Espe'*'. The re-activation of 
a saturated sorbent can be carried out by allowing the pump to warm 
to room temperature, care being taken to vent the pump to avoid a 



Valve closed 



To system 

Valve open 



Vent — 



Pump !■ 
(re-activating) 




Neoprene 
bung 

^Liquid' 
nitrogen 



Pump 2 
(pumping) 



Figure 3.16. Two sorption pumps for continuous pumping 



possible explosion caused by the pressure of desorbed gas. The 
re-activation process can be made more efficient by heating the 
pump whilst removing the desorbed gas by a mechanical pump. 

Table 3. 3. Characteristics of Some Sorption Pumps 



Pump 


MSSIO* MSS50* 

i 


MSSlOO* 


Max. volume (1.) pumped from atmospheric 
pressure to ultimate pressure of: 

IxW'tOTT 
3x lO-'tOTT 


1 
2 


8 
15 


15 
30 


Amount (g) of absorbent 


50 


250 


500 


Time (min) to pump to 10'^ torr 


10 


16 


20 


Coolant 


Liquid nitrogen at — 196°C 



• Vacuum Generators Ltd. 



61 



THE PRODUCTION OF VACUUM 

In order to obtain continuous sorption pumping (for example, 
when backing a vapour pump) it is necessary to use two pumps 
alternately, one pumping whilst the other is valved off from the system 
and is being desorbed. A typical arrangement for this is illustrated 
in Figure 3.16. 

The characteristics of some commercially available pumps are 
summarized in Table 3.3. In order to obtain shorter pump-down 
times, or for pumping larger volumes, several pumps may be used in 
parallel. 



3.6 Cryogenic Pumps 

Cryogenic pumping depends upon the fact that if a surface within a 
system is cooled then gases and vapours will tend to condense upon it, 
thus reducing the pressure. It can be seen by reference to vapour 
pressure data*®' that if the surface is cooled by liquid helium then all 
gases except helium will be condensed upon it. If liquid helium 
boiling under reduced pressure is used then even helium may be 
condensed. 

The ultimate pressure of such a pump for a given gas is determined 
by the vapour pressure at the temperature of the condenser surface 
of the condensed gas. Suppose the gas has a vapour pressure />„ at the 
temperature T„ of the condenser surface. It is easy to show*^' that 
the pressure in the system at temperature T (in this case the ultimate 
pressure for the particular gas) is given by 



--'-tm 



(3.6) 



where a and a„ are condensation coefficients {i.e. the probability that a 
molecule striking a surface is condensed) at temperatures T and 7\, 
respectively. Experimentally both a and a„ are close to unity, and 
hence for condenser surfaces at liquid helium and liquid hydrogen 
temperatures (r„ = 4-2°K and 20-4°K respectively) the values of 
p^jp^ are 84 and 3-8, assuming the bulk of the pump is at room 
temperature. Total ultimate pressures of the order 10~^° are readily 
attainable, using liquid helium boiling under reduced pressure. 

To estimate the speed of such a pump, consider an area A cm^ of 
condenser surface. This may be regarded as an aperture of area A 
era? with gas at pressures p^ and p on either side of it. From eqn 1.11, 
and taking condensation coefficients into account, the number of 
molecules per second leaving the regions at p and p„ respectively, are 

62 



I 



CRYOGENIC PUMPS 



AN 

-^ = apA{2nmkT)--'-i^ 



and 



dA^ 
dt 



a^p^{27rmkT^)-^i' 



Multiplying each of these by kT and taking their difference leads to 
the net rate of flow q of gas onto the condenser, measured at temperature 
T, as 



/ kTV^r /TV^'] 



Eliminating a„/)„ by eqn 3.6 and dividing hy p gives an expression for 
the pumping speed as 

p \2W \ p; 

This may be written in the form 



where 



5 = 5,(l-^)l.sec-i 
S^ = 3-64a^(^j 1. se( 



(3.7) 
(3.7a) 



Thus, the speed/pressure characteristic for cryopumping of a given 
gas has the same form as that of a rotary pump (see eqn 3.3 and Figure 
3.8). Since S is dependent upon both M and /»„ (because oi p^ the 
pump will, in general, have significantly different speeds for different 
gases. Provided that a is independent of the condenser temperature, 
then so is the value of S. 

A pumping action known as cryotrapping has been observed"' 
in addition to normal condensation. This process results in the appa- 
rent condensation of non-condensable gases (for example, the pump- 
ing of nitrogen in the presence of water vapour by a condenser at 
77°K, where the s.v.p. of nitrogen is 760 torr). It is believed to be 
due to the non-condensable gas being carried down by a condensable 
vapour and trapped within the pore structure of the condensate. 

A method of pumping gases which are non-condensable at the 
condenser temperature used, is to combine sorption pumping with 
cryopumping. An example of this is the system described by Cald- 
wooD, Gareis and Simson'^' in which a molecular sieve material is 
bonded to the condenser surface. Using liquid helium at 4-2°K, 

63 



THE PRODUCTION OF VACUUM 

ultimate pressures of the order of 10"*^ torr with pumping speeds of 
1,300 1. sec~^ and 7,900 1. sec"^ for hydrogen and helium respectively 
are reported 

A limitation upon the pumping efficiency at high pressures is 
imposed by the layer of solids which accumulates on the condenser 
surface. These solids (frozen gases and vapours) are usually of low 
thermal conductivity and so produce a significant increase of tempera- 
ture at the solid surface compared with the temperature of the 
condenser surface. The rate of build-up of solids is high at high 
pressures, being (typically) of the order of 10 cmh"-^ at 10"^ torr and 
10-2 (,jn h-i at 10"* torr. 



Condenser 
coils 

Temperature 
sensing element 

Throttle valve - 




=^ 



-Gas flow 



Gas pump' 



Jl 



Z^ZEr--'^° roughing 
pumps 



| — Liqu 



id flow 



IcXD 



-Vacuum insulated- 
feed tube 



.Liquid gas- 
reservoir 



Figure 3.17. Schematic representation of a cryopump 

A practical cryogenic pump is illustrated in Figure 3.17. It consists 
of a helix made of stainless steel tube, which acts as the condenser 
surface, mounted directly in the chamber to be evacuated. The 
coolant (liquid hydrogen or helium) is supplied to the helix via a 
vacuum insulated feed- tube from a Dewar storage vessel, and is made 
to flow through the spiral by means of a gas pump at the outlet end of 
the spiral. The coolant boils as it passes through the helix taking the 
necessary latent heat from the tube, which is hence cooled, and the 
resulting gas is drawn off by the pump. The rate at which coolant 
passes through the system, and hence the temperature to which the 

64 



ION PUMPS 

condenser surface is cooled, is controlled by a throttie valve mounted 
in the gas exhaust line. A temperature sensing element mounted on 
the condenser coils automatically controls the throttle valve setting. 
A metal plate cooled to liquid nitrogen temperature is often mounted 
above the condenser to act as a shield preventing excessive heating 
of the condenser by radiation from the vessel walls. Although the 
shield provides a baffling action which reduces the pumping speed, it 
is necessary to avoid excessive waste of coolant. 

The cryogenic pump is not used at pressures above 10 "^ torr, 
pardy because of the large quantities of coolant that would be required 
to pump directly from atmospheric pressure, and partly because the 
thickness of solid built up during high pressure pumping would seri- 
ously reduce the pump efficiency at low pressures. 

Pumps having speeds for nitrogen up to 5,000 1. sec" -^ (liquid 
helium consumption rate of the order of 2 1. h"-"^) are commercially 
available. Pumps with speeds up to 10^ 1. sec"-' are feasible, but for 
these high speeds the coolant would be fed directly from a gas lique- 
faction plant rather than from a storage vessel. It would also be 
more economic to use a subsidiary condenser at the temperature of 
liquid nitrogen to remove gases condensable at that temperature, 
before bringing into operation the liquid helium condenser. 

3.7 Ion Pumps 

If a gas is ionized and the resulting positive ions are attracted to a 
negatively charged plate, atoms of the gas are effectively removed 
from the system and thus a pumping action is produced. Both the 
hot and cold cathode ion gauges described in Chapter 4 act as pumps 
in this manner. Since the proportion of the gas ionized in such devices 
is small the rate of pressure reduction, and hence the pumping speed, 
is very small. Pumping speeds for hot cathode ion pumps are 
typically of the order of 10" '^ 1. sec"-"^, but can be up to 5 1. sec"-*^ for 
cold cathode (Penning) pumps. Such pumps are normally only 
used on small closed systems which have been previously evacuated to 
about 10 "'^ torr. 

An advantage of any pump based on the ion pump principle is that 
it incorporates its own pressure measuring device. 



3.8 Getter Pumps 

The gettering process described in Section 1.13.1 can obviously be 
used to remove gas from a system and hence produce a pumping 
action. However, the gettering of inert gases is relatively ineffective 

5~1.H.V.T. 65 



THE PRODUCTION OF VACUUM 

since it depends only upon physical sorption. For this reason getters 
are not generally used alone, but are combined with ion pumping to 
ensure the removal of inert gases. Such pumps, known generally as 
getter-ion pumps, may be discussed under the headings of (i) those 
using hot cathode ionization (evapor-ion pumps), and [ii) those using 
cold cathode ionization (usually employing sputtering and hence 
called sputter-ion pumps) . 

3.9 EvAPOR-ioN Pumps 

These pumps are of two basic types : 

{a) Small, low speed pumps designed to perform a single pumping 
operation and then to be discarded. 

{b) Large high speed pumps which can pump more or less con- 
tinuously with intermittent maintenance. 



Wire feed spool 



Titanium wire 



Heated post 
( + 1,000V) 

Accelerator 
grid (+1,000 V) 




Spool shaft, driven by 
external motor 



Wire guide 
Cooling water coils 



Filament (+100V) 



Figure 3.18. An evapor-ion pump 



The first type consists typically of a hot cathode ionization gauge of 
the Bayard-Alpert type (see Section 4.9.1 and Figure 4.16) containing 
an extra filament around which is wrapped a wire of getter material 
(usually titanium or zirconium). Both ion pumping and gettering 
(by electrically heating the getter filament) may be carried out at 



66 



SPUTTER-ION PUMPS 

pressures below 10"^torr. A pin in the glass wall enables the 
deposited getter film to be maintained at a negative potential thus 
helping to trap positive ions. The steady build-up of getter deposit 
on the pump walls and on the normal cathode buries the positive 
ions and ensures that when they are neutralized they are not sput- 
tered off by succeeding ions. Obviously when all the getter has been 
evaporated, only ion pumping and contact gettering can take place. 

A pump typical of the second type is illustrated in Figure 3.1 Shy 3. 
cut-away view, based on a pump manufactured by Consolidated 
Electrodynamic Corporation. A spool, carrying titanium wire, is 
externally controlled so that the wire is fed downwards onto a post 
of refractory conducting material maintained at 1,000 V positive with 
respect to the pump wall. Electrons produced at the circular filament 
(100 V positive with respect to the wall) bombard the post and heat 
it to about 2,000°C. This causes rapid evaporation of the titanium 
wire which then condenses on the cooled pump walls. The con- 
tinuous evaporation of the wire ensures a continuous pumping action 
both by dispersal and contact gettering. 

A wire mesh grid, at a potential of 1,000 V positive with respect to 
the walls, also attracts electrons from the filament and these cause 
ionization of the gas with the positive ions travelling to, and being 
largely retained by, the pump walls. Using a titanium evaporation 
rate of about 5 mg min~^, pumping speeds are of the order of 
3,000 1. sec-i for hydrogen, 2,000 1. sec-^ for nitrogen, 1,000 1. sec"! 
for oxygen, and 5 1. sec"-^ for argon. 



3.10 Sputter-ion Pumps 

A pump of this type is illustrated in Figure 3.19, and consists essen- 
tially of a stainless steel vessel containing an anode of honeycomb 
construction. Opposite each open end of the anode is mounted a 
titanium plate which acts as a cathode. A potential of about 3,000 V 
is maintained between the electrodes and a magnetic field of about 
1,500 G is applied by external permanent magnets along the axis of 
the electrode system. Positive ions of system gas which are formed in 
the region of the electrodes are accelerated to the cathode and acquire 
sufficient energy to sputter titanium. 

The sputtered titanium condenses mainly on the open structiu-e 
anode and in so doing ptmips active gases by both dispersal and con- 
tact gettering. Gas molecules which reach the anode by either of 
these processes are rapidly buried beneath succeeding layers of 
titanium and are thus permanently removed from the system. On 
the other hand, gas which reaches the cathode as positive ions has a 

67 



THE PRODUCTION OF VACUUM 

high probability of being desorbed by succeeding ion bombardment. 
This is particularly so in the case of the inert gases since they can only 
be ion pumped and then held at the cathode by the relatively weak 
forces of physical adsorption. 

Of the inert gases, argon is the most troublesome in this respect and 
is the main factor governing the ultimate pressure attainable. The 
problem of argon re-emission can be overcome to some extent by 
using an additional pair of cathodes set outside the main (sputter) 
cathodes which are perforated. The subsidiary (pumping) cathodes 



Titanium cathodes 




Figure 3.19. 



Stainless steel 
anode 



Direction of 
magnetic field 

The electrode structure of a sputter-ion pump 



are operated at a potential between that of the main cathodes and the 
anode. Many of the positive ions pass through the main cathode 
plates and are then decelerated towards the subsidiary cathodes which 
they strike with insufficient energy to cause gas re-emission. The 
effect of this modification, for a given set of conditions, is to improve 
the pumping of argon but to reduce the pumping speed for other gases 
because of the reduced sputtering rate. With such a pump, ultimate 
pressures of 5 x 10~^° torr have been reported. 

A further disadvantage is the care which must be taken when start- 
ing the pump at high pressures (max. 10"^ torr). The ion current at 
high pressures is large and causes heating of the pimip. If the pump 
has previously handled much gas the temperature rise leads to out- 
gassing which, in turn, causes a larger ion current. Such a process 
rapidly becomes 'run-away' leading to glow discharge between the 
electrodes and a rapid rise in system pressure. Even if the gas 

68 



REFERENCES 

evolution is not rapid pumping speed is reduced giving what is termed 
'slow starting'. These troubles can be largely overcome by initially 
pumping to pressures of the order of 5 x 10"* before operating the 
sputter-ion pump. 

The life of a sputter-ion pump is limited by the fact that eventually 
all the titanium becomes gas saturated. Saturation obviously occurs 
much more quickly if the pump is used continually at a high pressure. 
Commercial pumps are usually quoted as having a life of up to 50,000 h 



Is 6 

'" 5 











Slow 
starting" 






Hydrogen 
















































Nitrogen 




















Oxygen , 






' 








\ 










\ 






Argon " 












■>.^ 



2^zz 



1 


10" 



10"' 



10"' 10"^ 

Pressure, torr 



10-' 



10" 



Figure 3. 20. Typical characteristics of a sputter-ion pump with 
nominal speed 5 1. sec " ^ 



at a pressure of 10 ~® torr, and are available in speeds ranging from 
1 to 400 1. sec"^, but there appears to be no reason why pumps of 
much higher speed cannot be produced. 

A typical characteristic of a sputter-ion pump, showing the selective 
gas pumping action, is illustrated in Figure 3.20. 



REFERENCES 

1 Power, B. D. and Kenna, R. A. Vacuum 5 (1955) 35 

= Read, P. L. Vacuum 13 (1963) 271 

^ Hersch, C. K. Molecular Sieves. Rheinhold, New York, 1961 

* EsPE, W. Explle Tech. Phys. 12 (1964) 293 

5 HoNiG, R. E. and Hook, H. O. RCA Rev. 21 (1960) 360 

69 



THE PRODUCTION OF VACUUM 

Kennard, E. H. Kinetic Theory of Gases, p 66. McGraw-Hill, New York and 

London, 1938 
ScHMiDLiN, F. W., Nefliger, L. O. and Garwin, E. L. Vacuum Symp. Trans. 

(1963) 197 
Caldwood, R. R., Gareis, P. J. and Simson, J. P. 12th Natn. Symp. Am. 

vacuum Soc. (1965) Paper 10.4 



70 



MEASUREMENT OF LOW PRESSURE 



4.1 Introduction 

In the majority of vacuum systems a measurement of the pressure at 
various points is required, and a wide variety of gauges is available 
for this purpose. This chapter is intended to give an outline of the 
more important of these gauges. 

4.2 Manometer 

The manometer is one of the simplest and most fundamental methods 
for measuring pressure differences and consists of a U-tube containing 



Atmosphere 




Liquid of 
"density p 



Figure 4.1. Simple manometer 



liquid connected between the low pressure and the atmosphere (see 
Figure 4.1). The levels of the liquid in the two arms of the U-tube 
differ by a height / mm and hence the pressure difference between the 

71 



MEASUREMENT OF LOW PRESSURE 

atmosphere and the vacuum is YoPgl dyn cm "2, where p is the density 
of the liquid and g the acceleration due to gravity. If the liquid is 
mercury and / is measured in millimetres, the pressure difference is 
approximately / torr. The actual pressure in the system is thus 
equal to the atmospheric pressure (in torr) minus /. 

The determination of pressure t)y the manometer depends on a 
knowledge of atmospheric pressure, which is a variable quantity. 
One method of avoiding this is to use a differential manometer in 
which one arm is connected to a vessel at a pressure very much lower 
than that to be measured. The gas pressure in the system then sup- 
ports a height of liquid /' in the other limb and again, if the liquid is 
mercury, the pressure in the system may be taken for all practical 
purposes as /' torr. 

When the actual pressure in the system falls below 1 torr is it difficult 
to make accurate readings with either form of manometer. The 
sensitivity of the differential manometer may be improved by using a 
liquid (such as oil of low vapour pressure) which has a much lower 
density than mercury. In order to calculate the pressure difference in 
torr, / must then be multiplied by the ratio p ojl mercury Fo"" 
diffusion pump oil, p^n = 0-9 and the ratio is approximately 0-07, 
so that pressures down to about 0-1 torr can be measured in this way. 

Carr'^' describes a modified form of manometer which can 
measure pressure differences down to 1 " ^ torr. The two arms of the 
manometer are made very wide (about 2 in. in diameter) so that the 
mercury in the centre of the tube has a flat surface. The mercury 
levels are measured by means of sharp pointers attached to micro- 
meter heads, the pointers being adjusted so that they just touch the 
mercury surfaces. Carr also discusses other methods of measuring 
the height of the mercury, including the use of interferometry and 
electrical capacitance. 

4.3 Bourdon Gauge 

The Bourdon gauge consists of a spiral coil of hollow tubing of 
elliptical cross-section sealed at one end and connected at the other 
to the low pressure to be measured (see Figure 4.2). If the pressure 
inside the tubing increases, the tube cross-section tends to become 
more circular and this in turn causes the radius of the spiral to in- 
crease ; a reduction in pressure causes the spiral to decrease in radius. 
A pointer is attached by a mechanical linkage to the free sealed end of 
the spiral and moves over a scale calibrated in torr. The Bourdon 
gauge finds most use as a high pressure gauge but can be used to 
measure rough vacua down to about 10 or 20 torr. 

72 



aneroid capsule 

The readings indicated by this instrument are again dependent on 
the pressure difference between the inside and outside of the tube, 
and variations in atmospheric pressure, which may be as much as 40 
torr, cause errors which are particularly serious at the lower end of the 
scale. The exact relation between scale reading and pressure dif- 
ference is difficult to determine theoretically and therefore the gauge 
is calibrated against a mercury manometer. 




1 Vacuum system 
Figure 4.2. Bourdon gauge 



4.4 Aneroid Capsule 

Another gauge which depends on a mechanical system is the aneroid 
capsule, where the mode of operation is similar to that of an aneroid 
barometer. The capsule is in the shape of a sealed cylinder with thin 
corrugated ends, containing air at a fixed low pressure {see Figure 4.3). 
When the external pressure varies, the forces acting on the ends of the 
cylinder cause a small mechanical displacement which is magnified by 
a lever system and transferred to a pointer moving over a scale. As 
the enclosed gas is at a fixed pressure, provided its temperature re- 
mains constant the aneroid capsule is unaffected by changes in atmos- 
pheric pressure and is said to be barometrically compensated. 

The range of the gauge may be altered by using a different pressure 
inside the capsule, typical ranges being 760-0 torr, 20-0 torr and 
10-0 torr. The lowest pressure measurable by the aneroid capsule is 
about 1 torr. 

73 



MEASUREMENT OF LOW PRESSURE 



Again this gauge has the disadvantage of needing to be calibrated 
against a manometer; on the other hand it provides a convenient 
and robust measuring instrument for rough vacua. 



Mechanical linkage 




Figure 4.3. Aneroid capsule gauge 



4.5 Discharge Tube 



The colour and form of the electrical discharge in a gas at low pressure 
may be used to estimate pressures in the range 5-0-01 torr. 



5 torr 



1 torr 



^£i ^Isi^ 



Spark 
5x10"' torr 



Discharge of characteristic colour 
10^^ torr + 



- jm I ! Ml J in![> - (pmm \ iTl) 



Striations 



10"2 torr 



^ 



Fluorescence- 



Figure 4.4. Discharge tube 



The discharge tube is an elementary form of ionization gauge in 
which a potential difference of several thousand volts is applied between 
two electrodes in a narrow glass tube connected to the vacuum 
system. An electron in the tube is accelerated by the electric field 

74 



MCLEOD GAUGE 

and, provided its mean free path is sufficiently long, it gains enough 
energy to cause ionization and excitation of the gas molecules. The 
ionization results in a further supply of electrons to maintain the dis- 
charge. The excitation energy of the gas molecules is lost by the 
emission of light, the colour of which is characteristic of the gas in the 
tube. In the case of air, a pink discharge is produced. As the pres- 
sure falls the form of the discharge changes and Figure 4.4 shows the 
approximate correlation between the form of the discharge and the 
pressure. It can be seen that the form of the discharge is not par- 
ticularly sensitive to pressure changes. 

When the pressure reaches about 10 ~^ torr the number of collisions 
is not sufficient to maintain an easily visible discharge. The elec- 
trons, however, bombard the walls of the tube and fluorescence of the 
glass may be observed. As the pressure continues to fall the supply of 
electrons decreases further, hence the fluorescence becomes less and 
disappears completely at about 10"^ torr. This condition is known as 
black-out. 

The discharge tube is particularly useful in leak detection and this 
application is discussed more fully in Chapter 7. 



4.6 McLeod Gauge 

The McLeod gauge'^' is an extremely important instrument, being the 
most frequently used absolute gauge over the range 10-10"^ torr, 
although below 10"* torr the accuracy is limited as described on 
pp. 78-80. It is a modification of the manometer and its construction 
is illustrated in Figure 4.5. The glass bulb A is attached to a narrow 
capillary tube B which is closed with a flat top. A is also connected, 
via the Y-joint at C, to both the evacuated system and the mercury 
reservoir. The tube D to the vacuum system has a parallel side 
branch E, made of capillary tubing of the same diameter as B, which 
ensures that when both tubes contain mercury the depresssions of the 
menisci due to surface tension are equal and hence may be neglected. 
E and B are arranged to be adjacent to one another to facilitate 
comparison of the mercury levels. 

Consider the mercury level to be below C and the whole system at a 
pressure of/» torr. Let the total volume of ^ and B above the level 
X-X be V mm^, and the radius of 5 be ?• mm. If the mercury is 
raised to X-X, a voliune of gas V at pressure p is isolated in A and 
B. The mercury level is then raised slowly until it reaches a mark on 
E exactly opposite the closed end of J5. During this process the gas in 
A and B is compressed by the head of mercury in E and finally occu- 
pies a length / of the capillary tube as indicated in the diagram. 

75 



MEASUREMENT OF LOW PRESSURE 



The new volume of gas is thus nr'^l and the total pressure acting on 
it is the sum of p and that due to the column of mercury of length /. 
If /is measured in millimetres the total pressure is {l+p) torr. As the 
compression takes place slowly the temperature may be assumed to 
remain constant, so that Boyle's Law can be applied. Thus 



pV = 7rrH{l+p) 



In general p <!^ I and so 



P = 



irrH^ 



(4.1) 



If the dimensions V and r are known the gauge may be calibrated 
absolutely to measure pressure. It is the only gauge in common use 




Figure 4.5. McLeod gauge 



which can be calibrated in this way, and for this reason the McLeod 
gauge is extremely important. The calibration of the gauge is 
independent of the nature of the gas in the system provided the gas 
obeys Boyle's Law. 

In the mode of use described above, it should be noted that the 

76 



MCLEOD GAUGE 

pressure scale is not linear. An alternative but less common method 
of operating the McLeod gauge which produces a linear relationship 
between pressure and length, is to bring the mercury level up to a 
fixed mark on the closed capillary B a distance l^ from the closed end 
so that the compressed gas always occupies the same volume tttH^. 
The difference in heights 4 of the two mercury columns is now read on 
E and the effective pressure in torr acting on the gas in B is {p + l-^. 
Applying Boyle's Law again gives 

pV = nr\{p+l2) 

As li, r and V are constant p is directly proportional to /g. 

4.6.1 Sensitivity and Range 

The sensitivity of the McLeod gauge may be defined as dljdp, 
the change in height of mercury for unit change in pressure. In the 
first mode of use described above 

d/ V 

dp ~ 2TTrH 

The sensitivity is thus increased as Fis increased and r decreased. The 
sensitivity is also a function of I, indicating different sensitivities at 
different points on the pressure scale, the maximum sensitivity 
occurring at small values of /, i.e. at low pressures. For the second 
mode 

d/a _ V 

dp 7Tr\ 

This is independent of ^2 and thus constant at all points over the range. 
The sensitivity may however be increased by making [^ small. The 
ultimate limitation on sensitivity is determined by the practical limits 
to the values of V, r, l^ and /. For example, if r < 0-5 mm it is found 
that surface tension forces hold the mercury thread in the capillary tube 
even when the mercury in the bulb A is lowered. Also, if V is very 
large the quantity of mercury required to fill the gauge is excessive, 
and in addition the weight of mercury tends to distort the glass and 
thus falsify the readings. The maximum practicable value of V is 
about 5x10^ mm^. The final limit on sensitivity is set by the diffi- 
culty in measuring the lengths / or /j accurately when they are less 
than about 1 mm. Thus the sensitivity of a McLeod gauge which has 
r = 0-5 mm, F = 5 x 10^ mm^ and I or l^ = \ mm is 6-3 x 10^ mm 
torr"^ in the linear mode and 3-1 x 10^ mm torr"^ in the square 

77 



MEASUREMENT OF LOW PRESSURE 

law mode. The majority of gauges in ordinary use have lower 
sensitivities than this; for example, a typical gauge has F = 
1x10^ mm^, r = 1 mm and the sensitivity at / = 10 mm is thus 1 -6 x 
10^ mm torr"^. 

Another important feature which is of considerable interest to the 
user is the range of pressures which a given McLeod gauge is capable 
of measuring. The parameters which determine the range are also r, 
V, the minimum measurable value of I and, in addition, the total 
length of the capillary tube. Practical considerations limit the 
capillary tube length to about 100 mm and the limitations on r, V 
and / are the same as for the gauge sensitivity. 

As an example, consider a gauge where F = 10^ mm'^ and r = 1 mm. 
Substitution in eqn 4. 1 for the square law form of operation shows that 
when I = 100 mm then/) = 3 x 10~^ torr, and when / = 1 mm then 
p = 3x 10"® torr. The range of this gauge is thus 3 x 10~^-3 x 10~® 
torr {i.e. four decades). On the other hand if the gauge is operated 
in the linear form, it can be seen that if /^ =1 mm when l^ = 100 mm 
then/) = 3 X 10"^ torr, and when /g = 1 mm then p = 3 x 10~° torr 
{i.e. only two decades). The range in this mode can, however, be 
increased by using values of /i of 1 or 1 00 mm. Thus for /^ = 100 mm 
and I2 = 100 mm, p = 3 x 10 "-^ torr and the total range is the same 
as in the square law mode. 

4.6.2 The Effect of Condensable Vapours 

One serious fault of the McLeod gauge is that it does not measure 
accurately the contribution to the total pressure of any vapours in 
the system. The McLeod gauge is only accurate for gases obeying 
Boyle's Law but the discussion on vapours (Section 1.9) indicates 
that after saturation a vapour condenses and then exerts its s.v.p., 
irrespective of the volume of the vessel. This situation is likely to 
arise inside the capillary tube of the McLeod gauge when the gas 
and vapour are considerably compressed from their initial volume. 

If the vapour does not become saturated during compression, it is a 
reasonable approximation that its behaviour will be governed by 
Boyle's Law and hence the gauge will read the total pressure due to 
both gas and vapour. On the other hand, if condensation does take 
place, the gauge will read the pressure due to the gas plus the s.v.p. 
due to the condensed vapour. In this case, even if the s.v.p. at the 
temperature of measurement is known, it will be impossible to calcu- 
late back to the original total pressure. In any case, it is difficult to 
determine whether condensation has taken place as the quantity of 
liquid produced will be extremely small and not easily visible. 

The most common vapour present in a vacuum system is that of 

78 



MCLEOD GAUGE 

water. Let p„ be the partial pressure in the vacuum system due to 
water vapour. In the McLeod gauge the initial volume V of water 
vapour at pressure /)„ is compressed to occupy a volume -nrH in the 
capillary tube at a pressure p!„. Assuming that water vapour obeys 
Boyle's Law down to its condensation point, then p'„ = p^Vj-n-rH 
and condensation occurs if/)^ ^ s.v.p. of water vapour at the tempera- 
ture of the gauge. Similarly, the minimum value of the head of 
mercury / which can cause condensation is also numerically equal to 
the value of the s.v.p. in torr. This would be the case with only water 
vapour in the system, a situation unlikely to arise in practice. 

Thus, if /)„' and / are set equal to s.v.p., the minimum partial pres- 
sure of water vapour which can condense on compression is equal 
to (s.v.p.)^X7rr^/F. For a gauge at 20°C where r = \ mm and 
V = 10® mm^ this pressure is equal to about 10"^ torr (s.v.p. of water 
vapour =17 torr at 20°C) . The significance of this result is that, for 
this particular gauge, any water vapour in the system exerting a partial 
pressure of less than 10"^ torr will be recorded more or less correctly. 
In general, however, the initial concentration of water vapour is less 
than 100 per cent and thus the critical value of/)„ for condensation to 
occur is greater than 10 ~^ torr. The calculation must be performed 
for each individual gauge, but it can be seen that the greater the value 
of V or the smaller the value of r, the lower is the water vapour pres- 
sure which may result in condensation. The maximimi practical 
value of F = 500 cm'^ and the minimum value of r = 0-5 mm, leading 
to a minimum critical value for/)„ of 4 x 10"* torr. 

These minimum critical values of /)„ occur within the range of 
pressures produced by normal rotary backing pumps. If no gas ballast 
is used and no drying trap employed, the concentration of water vapour 
in the vacuum system could become quite high and under these 
conditions the errors in the McLeod gauge readings would be serious. 
If a McLeod gauge is used in such a system it is therefore essential to 
use a drying trap so that the measured pressure is that due to the air 
only. On the other hand, at lower pressures such as those produced 
by a diffusion pump, the McLeod gauge appears to give reasonably 
accurate pressure readings without the danger of condensation. 

If vapours other than water are likely to be present in the system a 
similar calculation to that above will indicate the limit of accurate 
pressure measurements of the McLeod gauge. 



4.6.3 The Effect of Outgassing 

A further limitation on the use of the McLeod gauge is set by the 
outgassing of the gauge bulb and also of the connecting pipe. By its 
mode of construction, it is usual for the McLeod gauge to be connected 

79 



MEASUREMENT OF LOW PRESSURE 

to the vacuum system by a long pipe having a large surface area and 
at the same time a small conductance. The desorbed gas flowing 
down the connecting pipe sets up a pressure difference between the 
system and the gauge head, so that pressures measured at the gauge 
head are too high. 

An estimate of this pressure difference may be made as follows. 
The rate of desorption of gas by the bulb = 4:TTrfK, where r^ is the 
equivalent radius of the bulb and K the outgassing constant. If the 
pipeline is of length L cm and diameter D cm, its molecular flow 
conductance is 12-1 D^jL 1. sec"^ and the pressure drop d/)i across it is 
AnrlKLj 1 2 • 1 D^. Similarly, the rate of evolution of gas from the pipe 
is ttDLK; it may be assumed that the whole of this originates at the 
centre of the tube"' and hence the pressure drop dp^ due to this gas is 
ttDLK X Z,/24-2Z)3. The total pressure drop is given by the sum of 
dpi and dp2, i.e. 

tttIKL ■nKL\ 



3/)3 -> 24-21)2 

In a case where ri = 3 cm, £» = 1 cm, Z, = 100 cm and K = 10"'' 
torr 1. sec~^ cm"^, the total pressure drop = 2-5 x 10~* torr. Thus 
the pressure indicated by this gauge is always too high by 2 -5 x 1 " * torr 
and measurements of pressures less than 10"^ torr are seriously in 
error. The means of reducing this error are to reduce K by degassing 
the bulb and pipe and to ensure that the conductance of the pipe is as 
large as possible. 

Another source of error is due to the cold trap which is normally 
interposed between the gauge head and the system to prevent mercury 
vapour entering the system. The trap acts as a mercury condenser 
causing a steady stream of mercury vapour to flow from the gauge to 
the trap. This streaming creates a pumping action similar to that in a 
diffusion pump, producing a lower pressure within the gauge than 
that in the system. This phenomenon has been carefully investigated 
and the work of Ishii and Nakayama'*' may be referred to. 

4.6.4 Practical Forms of the McLeod Gauge 

There are in use several commercial forms of the McLeod gauge 
and a few of these are now described. 

4.6.4.1 Mercury Reservoir Type— The mercury to fill the gauge is 
contained in a reservoir which can be raised or lowered and which is 
connected to the vacuum system via the Y-joint C. Atmospheric 
pressure acting on the mercury in the reservoir causes it to rise in the 
tube towards C to a height of about 75 cm. In order to take a reading 

80 



MCLEOD GAUGE 

the reservoir is raised from below the Y-joint and the level of mercury 
in the gauge also rises and can be adjusted to the appropriate level 
in E. For the next reading the reservoir is lowered until the bulb B 
is again in communication with the vacuum system. This type of 
gauge must therefore have a stem at least 75 cm long to accommodate 
the atmospheric head of mercury and is consequently rather cumber- 
some. 

4.6.4.2 Bench Model — ^The overall height of the McLeod gauge can be 
reduced by using a subsidiary pump to raise and lower the mercury. 
Figure 4.6 shows a typical arrangement. To raise the mercury, air is 



Vacuum 
system 




Figure 4.6. Bench McLeod gauge 






Subsidiary 
vacuum 



admitted to the reservoir by suitable adjustment of the two-way tap. 
Care is needed here because if the tap were fully opened the mercury 
would rise to a height of 75 cm and probably overflow into the pump- 
ing system. To lower the mercury the tap is turned to the subsidiary 
pump position. 

6 I.H.V.T. 8 1 



MEASUREMENT OF LOW PRESSURE 

4.6.4.3 Measuvac Type — This instrument is illustrated in Figure 4.7. 
The mercury (only a small volume of which is required) is contained 
in a flexible reservoir and its level is normally belovi? the lower ends of 
the capillary and reference tubes. In order to take a pressure reading 
the reservoir is mechanically compressed so raising the mercury into 
the tubes, isolating a sample of gas at system pressure, and compressing 
it as in the standard gauge. The reference tube also acts as the tube 
connecting the gauge to the system and is of relatively large diameter 



Vacuum 
system 




Figure 4. 7. Measuvac gauge 



to ensure a reasonably high pumping speed. Allowance for the dif- 
ferent surface tension depressions in the reference and capillary tubes 
is made by the manufacturer by setting the reference mark the ap- 
propriate distance above the closed ends of the capillaries. The use of 
two capillaries, respectively associated with different volumes, 
provides this instrument with a range 150-0-001 torr. This large 
range, combined with its compact and robust structure, makes it a 
very useful form of the McLeod gauge. 

4.6.4.4 Vacustat — ^The Vacustat is a compact but less sensitive form 
of the McLeod gauge and is illustrated in Figure 4. 8. The gauge head 
is mounted on a panel and can be rotated about its centre point. 
Because the volume Fis small and the capillary tube is of fairly wide 

82 



MCLEOD GAUGE 

bore, the gauge is only suitable for pressures in the range 1-10~^ 
torr. When placed in a horizontal position, the mercury flows into 
the reservoir and the rest of the gauge is evacuated ; on rotating to 
the vertical position the mercury rises into the measuring and reference 
tubes. The quantity of mercury used in the gauge is just sufficient to 
rise to the fixed mark on the reference tube at the lowest measurable 
pressure; slight tilting may be necessary at higher pressures. The 



Reservoir 





Figure 4.8. Vacustat gauge 



scale used is the square law one described above. Whilst not pos- 
sessing the range or accuracy of the normal McLeod gauge, the Vacu- 
stat is very convenient for making quick checks at backing pump 
pressures. 

4.6.4.5 Multirange McLeod Gauge'-^^ — ^The range of operation of a 
McLeod gauge may be extended by using the modified gauge shown 
in Figure 4.9. The single bulb of the ordinary gauge is replaced by a 
series of bulbs of volumes F4, Fg, V2 ending in a capillary tube and 
bulb B of volume V^. For very low pressures, the gauge is oper- 
ated on the normal square law principle, the mercury being brought 
to a fixed mark on the capillary tube E and the tube B being cali- 
brated as before to read pressures directly. 

In order to extend the range to higher pressures the gauge must be 
operated on the linear principle. Marks are provided separating 
each of the volumes Fj, Fg, Fg and F4 at points K, L, M and N. Thus, 
for the next pressure range, the mercury is raised to fill the gauge to 
point K and the height difference I between K and the mercury 

83 



MEASUREMENT OF LOW PRESSURE 



column in D is measured. (Tube D has the same diameter as the 
connecting tubes at K, L, M and N .) Applying Boyle's Law as 
before gives 

p{V^+V^+V^+V^) = {p + l)V, 



k Vacuum system 



Air 
Subsidiary 




Figure 4. 9. Multirange McLeod gauge 



Thus 



V,l 



A further extension of the range is made by raising the mercury to 
level L, whence 

{V, + V,)l 



P = 



84 



RADIOMETER GAUGE 

Finally, if the mercury is raised only to level M 

A suitable choice of the values Vi, V^, V^ and V^ will produce the 
ranges required. For example, a gauge for which V^ = 300 mm^. 
Fa = 19,700 mm3, Fg = 80,000 mm^ and V^ = 120,000 mm^ 
would cover the range of pressures 2 x 10 "^-100 torr. 

4.6.5 Advantages and Disadvantages of the McLeod Gauge 

The one great advantage of the McLeod gauge is that, if it is used 
carefully and correctly, it gives an absolute measurement of pressure. 
The McLeod is the only gauge in normal vacuum work which can be 
calibrated from its dimensions and which is independent of the 
nature of the gas in the system provided condensables are excluded. 
The accurate calibration of all other gauges can be performed against 
a McLeod. 

Against this important advantage must be set many disadvantages. 
The McLeod gauge is a large and awkward instrument which must 
be connected to the system by long pipes, and being made of glass the 
gauge head is fragile. The adjustment of the mercury levels is tedious 
and in some forms of the gauge, if care is not taken, it is easily possible 
for the mercury to overflow into the rest of the vacuum system. 
Another serious drawback is that the gauge is not continuously read- 
ing. 



4.7. Radiometer Gauge 

This gauge was first described by Knudsen*^' and may be used in the 
range 10 "^-10"^ torr, in which the mean free path is long compared 
with the vessel size. In theory, it can be caUbrated absolutely and 
should provide a useful standard of pressure measurement. How- 
ever, in practice it suffers from many disadvantages and is not in 
common use. For this reason only a brief description is given here ; 
a full account is given by Leck<''\ 

The Knudsen gauge operates on the radiometer principle and the 
basic construction of the instrument is illustrated in Figure 4.10. A 
light vane C is supported vertically at its centre point by a torsion 
wire, and two plates A and B heated to temperature T^ are arranged 
to lie parallel to the vane and on opposite sides. Gas molecules 
rebounding from A and 5 fall on areas jE and F of the vane and produce 
a force proportional to their momentum, i.e. to T^. At the same time 
gas molecules reflected from the walls of the vessel at temperature T^ 

85 



MEASUREMENT OF LOW PRESSURE 

fall on areas G and H of the vane and produce a force proportional to 
T'j. If 7^2 > 7\ there is a net couple on the vane and the resultant 
torsional twist in the suspension wire is measured by the conventional 
mirror, lamp and scale. The theory shows that the twist produced is a 
function of the pressure, the physical dimensions of the vane, and 
T2IT1. It does not depend on the nature of the gas and hence the 
gauge calibration should be absolute. 

The main disadvantages of the instrument are that (i) the simple 
theory does not hold in practice so the gauge cannot be calibrated 
absolutely, (n) the torsional suspension is fragile and susceptible to 



Suspension fibre 
Mirror 



Elevation 




Figure 4.10. Radiometer gauge 

vibrations, and {Hi) the working pressure range for a given torsion wire 
is small and several different wires are required to cover the whole 
range from 10"^ to 10~* torr. 

4.8 Thermal Conductivity Gauges 

4.8.1 Pirani Gauge 

The Pirani gauge '^' is one of the most widely used vacuum instru- 
ments in the range of medium high vacua. Its action depends upon 
the variation in thermal conductivity of a gas with pressure. This 
effect is discussed in Section 1 .8 where it is stated that at low pressures 
the thermal conductivity of a gas is a linear function of the pressure. 

Consider a metal filament heated by an electric current. The 

86 



THERMAL CONDUCTIVITY GAUGES 

wire reaches an equilibrium temperature at which the heat generated 
by the electric current is exactly balanced by the heat losses due to 
conduction, convection and radiation. Radiation losses are small 
unless the temperature is very high; at low pressures when the mean 
free path is comparable with the size of the vessel the distinction 
between convection and conduction disappears. Gas molecules on 
striking the heated filament gain kinetic energy which they transfer 
by collision either to other gas molecules or to the walls of the vessel. 
At high pressures the number of molecules striking the filament in 
unit time is large and the heat losses are also large. 



Figure 4.11. Pirani gauge head 



It should be noted that at high pressures, when heat transfer is all by 
intermolecular collisions, the heat losses are independent of the pressure. 
As the pressure falls a mixture of intermolecular and molecule-wall 
collisions is produced and the loss of heat is then some complex 
function of the pressure. As the pressure continues to fall the mole- 
cule-wall collision process becomes dominant and the loss of heat 
becomes directly proportional to the pressure. In general then, the 
heat losses become smaller as the pressure decreases and in consequence 
the temperature of the filament increases as the pressure falls. A rise 
in temperature also produces a rise in electrical resistance and a new 
equilibrium condition is established. Hence a measurement of the 
electrical resistance of the filament provides a means of determining 
the pressure. 

The pressure limits to the range of the Pirani gauge are set by the 

87 



MEASUREMENT OF LOW PRESSURE 



heat loss rates due to conduction and radiation. The heat loss due to 
radiation is small but independent of pressure. At the high pressure 
end ( ~ 500 torr) the heat loss due to conduction is also constant with 
pressure, and thus the reading of the gauge becomes independent of 
pressure. At about 10"^ torr the heat loss due to conduction 
becomes similar in magnitude to the heat loss due to radiation. 
Consequently, changes in pressure cause only minor alterations in 
heat losses and hence in filament resistance; the gauge becomes too 
insensitive for practical use. 

A typical Pirani gauge is illustrated in Figure 4.11. The nickel or 
tungsten wire is coiled into a spiral and held in the centre of the 
system by a wire support. Gauges with both glass and metal enve- 
lopes are available. 

Gauge ^ 



head - 




Figure 4.12. Pirani gauge control circuit 

The usual method of measuring electrical resistance is by some 
form of Wheatstone bridge and this is used with the Pirani gauge. 
Figure 4.12 shows the Wheatstone bridge circuit together with that 
of the power unit. Rj, is the resistance of the Pirani gauge filament, 
J?2 and i?4 are fixed resistances and R^ is an adjustable resistance. 
With the milliammeter connected in the VAC position, the balance 
condition is 



R, 
R,. 



Rs 
R, 



Rv = 



R2R3 
R. 



(4.2) 



One method of measuring the pressure is to balance the bridge by 
varying Rg and calculate the resistance of Rp, a previous calibration 
enabling this resistance to be converted into pressure. It is more 
usual, however, to keep R2 and R^ constant and preset R3, and to 
measure the out-of-balance current through G, which can be a 

88 



THERMAL CONDUCTTVITY GAUGES 

milliammeter whose scale is calibrated in pressure units. In this 
case it is essential to keep the voltage across the bridge constant. The 
bridge may be balanced initially at atmospheric pressure; then an 
increase in the resistance of Rp causes an increase in out-of-balance 
current so that the lowest pressures correspond to full-scale readings of 
the milliammeter. Alternatively, the bridge can be balanced at a 
fixed low pressure (< 10~^ torr) and then, as the pressure falls from 
atmospheric, the out-of-balance current decreases. In both these 
modes of operation the scale at the high pressure end is very compres- 
sed and becomes progressively more open as the pressure decreases. 
The normal form of the gauge is useful over a range of 500 x lO'^- 
Sx 10"^ torr. 

Commercial forms of the Pirani gauge are generally supplied with 
a control unit, a typical example of which is also illustrated in Figure 
4.12. The bridge voltage is supplied by a direct current derived from 
rectified a.c, and can be varied by means of a 'set voltage' rheostat. 
By switching to the position SV the millijimmeter can be used as a 
voltmeter and the voltage standardized against a 'set voltage' mark 
on the scale. In some instruments the bridge voltage varies with the 
pressure in the gauge head and standardization must be carried out 
before each pressure observation. 

Each control unit can be used with different heads of nominally 
the same type; in order to match exactly the resistances of the gauge 
heads to the bridge circuit (and hence to the scale calibration) it is 
necessary to have a supplementary rheostat in the Wheatstone bridge 
network. In Figure 4.12 this is the variable part of iJg and is adjusted 
by the manufacturer during calibration and then sealed in a small unit 
which is attached to the gauge head. 

It is not possible to calibrate the Pirani gauge from first principles. 
Although a theoretical value for the thermal conductivity of gas can be 
calculated there are a number of factors which make the direct applica- 
tion of this impossible. Firstly, over the range of pressures used, the 
mechanism of heat loss is a mixture of molecule-molecule and 
molecule-wall conduction; empirical methods have been used to 
estimate the relative importance of these two terms at different 
pressures but they are not very satisfactory. Secondly, the geometry 
of the gauge also has an influence on the heat losses. Another impor- 
tant factor is the nature of the gas in the system: the calibration of the 
Pirani gauge changes for different gases because of the different values 
of thermal conductivity. Thus, in a vacuum system pumping air the 
pressure indication will be made erroneous by the presence of vapours 
(especially that of water). However, for many purposes such an 
indication is adequate where only an order of magnitude of pressure is 

89 



MEASUREMENT OF LOW PRESSURE 

required. If a more accurate knowledge of the pressure is needed 
a suitable vapour trap must be used. 

There are a number of modifications by which a Pirani gauge may 
be made more accurate and also more sensitive. One of these 
concerns the variations in the resistance of the gauge head caused by 
ambient temperature fluctuations. In order to overcome this a 
dummy head is provided which has a resistance element of the same 
size and material as the active gauge head but is sealed in a glass 
envelope at a pressure much lower than 1 ~ ^ torr equal to that at which 
the bridge was balanced. The dummy head becomes the resistance 
/?2 (see Figure 4. 12) and ambient temperature changes then produce 
equal resistance changes in Rj, and R^. Thtis this source of error is 
eliminated. 

Another form of the Pirani gauge is the four-head model. In this, 
two arms of the Wheatstone bridge network {R^ and R^ are made 
variable and the other pair {R^ and R^ are kept fixed. i?2 ^nd R^ 
are dummy heads as described above. Rp and R^ are made into 
identical gauge filaments and both are placed in the vacuum system 
so that pressure changes affect them both equally, and hence ap- 
proximately double the out-of-balance current for a given pressure. 
This means of increasing the sensitivity of the bridge is particularly 
usefiil in the pressure range 5 x 10~^-1 x 10"^ torr where the sensiti- 
vity of the single-head type falls off. 

4.8.2 Semiconductor Gauge 

Instead of using metal filaments it is possible to construct a Pirani 
gauge from semiconducting materials (see for example Varicek*®'). 
The temperature coefficient of resistance of a semiconductor is 
negative but numerically much larger than that of a metal, and hence 
in a Wheatstone bridge circuit gives a much larger out-of-balance 
current at a given pressure. The circuit is basically the same as for 
the ordinary Pirani gauge, but it is essential to use a sealed compensat- 
ing thermistor in order to cancelout ambient temperature variations 
which could cause relatively large changes in resistance. The pres- 
sure range for the thermistor gauge is from 10"'' torr up to about 
10 torr, within which range the gauge calibration is almost linear with 
pressure. 

4.8.3 The Thermocouple Gauge 

A gauge operating on a similar principle to the Pirani gauge is the 
thermocouple gauge, in which a filament is heated electrically and its 
temperature measured directly by means of a thermocouple. As in 
the Pirani gauge, the temperature attained by the wire depends on the 

90 



IONIZATION GAUGE 

loss of heat by conduction and consequently the temperature reading is 
a measure of the pressure in the system. The filament operates 
at a temperature between 100° and 200°C and iron-constantan or 
Chromel-Alumel thermocouples have proved suitable. The thermo- 
electric e.m.f., which is only a few microvolts, can be used to operate 
a microammeter calibrated in pressure units. The operating range 
of the thermocouple gauge is from about 1 torr down to 1 ~ '^ torr. An 
advantage of this gauge is that the filament can be heated by an alter- 
nating current, thus simplifying the power supply. 

In general, the Pirani gauge has been preferred to the thermocouple 
gauge for vacuum practice. 

4.8.4 Summary 

The thermal conductivity gauge is one of the most widely used 
vacuum gauges over the pressure range 500 x 10~^-5 x 10"^ torr. It 
is a robust instrument and readings can be taken rapidly and continu- 
ously. On the disadvantage side, it requires calibration against an 
absolute instrument. 



4.9 Ionization Gauge 

The most commonly used gauge for measuring pressures below 10"^ 
torr is the ionization gauge, which exists in a number of forms all of 
which operate on the same basic principle. The residual gas in the 
gauge head is subjected to ionizing radiation and some of the gas 
molecules become ionized. The positive ions are attracted towards a 
negatively charged electrode placed nearby, and a very small electric 
current flows in an external circuit in order to maintain the charge 
on the electrode. The magnitude of the positive ion current depends 
on a number of factors including the intensity of the ionizing radia- 
tion, the nature of the gas and the number of gas molecules per unit 
volume, i.e. the pressure. For a given set of conditions the ion current 
is directly proportional to the pressure and the measurement of pres- 
sure becomes the measurement of this current. The ion current is 
only of the order of microamperes even at pressures as high as 10"'' 
torr, and at lower pressures amplification is necessary in order to make 
measurements possible. In theory at least, the gauge should be useful 
down to zero pressure, although in practice modifications are neces- 
sary for use below 1 ~ ^ torr. 



4.9.1 Hot Cathode Ionization Gauge'-^^^ 



An ionization gauge for use within the range 10 "''-lO"^ torr is 
own in Figure 4.13. The arrangement of the electrodes is similar to 

91 



MEASUREMENT OF LOW PRESSURE 

that in a triode valve. In the centre is a tungsten filament (cathode) 
and surrounding it is a helix of nickel wire (grid) . The ion collector, 
a cylinder of nickel, is concentric with the grid and filament. The 
grid is maintained at a positive potential of about 1 50 V with respect 
to the cathode whilst the collector has a negative potential with 
respect to the cathode of about 30 V. Electrons are emitted ther- 
mionically from the cathode when it is raised to a temperature of some 
2,000°G by the passage of a current of about 2 A. The electrons are 
accelerated by the positive potential on the grid and gain sufficient 
kinetic energy to cause ionization of the gas molecules present. The 



Collector 




Figure 4.13. Ionization gauge head 



Grid 



Filament 



open structure of the grid allows most of the electrons to pass through 
into the space between the grid and the collector and it is here that the 
majority of the ionizing collisions are made. 

The positive ions produced are attracted by the negative potential 
on the collector and the resulting current is measured by a microam- 
meter in the collector circuit. Figure 4.14 shows a typical electrical 
circuit for use with the ionization gauge. The voltages used in the 
gauge must be highly stabilized to prevent spurious variations in 
current, which would give apparent changes in pressure. 

In this connection it is particularly important to regulate the emis- 
sion of electrons from the cathode. This is almost invariably ac- 
complished by some form of feedback circuit such as that described by 
RiDENouR and Lampson*^". 

92 



IONIZATION GAUGE 

It is not possible to deduce an exact expression for the constant of 
proportionality between pressure and current and thus it is necessary 
to calibrate the gauge — usually against the McLeod gauge. Dif- 
ferent gases, because of their different ionization energies, produce 
different ion currents for a given molecular density and in Figure 4.15 
are shown a number of typical calibration curves for several gases. 
It should be noted that the ionization gauge records the presence of all 
gases and vapours. 

One of the potential sources of inaccuracy in this gauge is the gas 
evolved at the electrodes due to their heating by the electron and ion 
bombardment. Consequently, it is essential to outgas the grid and 




I 180V 



30V 



6A/20 (C]a) 

Figure 4. 14. Ionization gauge control circuit 

the collector thoroughly before each pressure reading is taken. In 
order to do this the grid and collector are connected together and made 
positive ( ~ 500 V) with respect to the cathode ; the electron current is 
increased by raising the filament temperature with the result that both 
electrodes are brought up to red heat, and the sorbed gases are then 
liberated and pumped away. 

When the ionization gauge is operated on a vacuum system using an 
oil diffusion pump the question of oil contamination of the filament 
arises. Oil or hydrocarbon vapour reacts chemically with the hot 
tungsten and the electron emission characteristics of the latter are 
changed. This in turn produces a change in ion current and an 
apparent change in pressure. It is therefore essential to employ 
a liquid nitrogen trap between the oil pump and the gauge so that 

93 



MEASUREMENT OF LOW PRESSURE 

this effect is minimized. It should be noted that similar effects may 
be produced with water vapour and some organic vapours. 

Spurious pressure readings may also be caused by surface films 
on the gauge envelope allowing small electrical currents to flow along 
the glass between the various electrodes. These currents can be 
greatly reduced by using an envelope construction in which the 
leakage paths between grid and collector are long and by ensuring 
that both the inner and outer surfaces between these electrodes are 
kept clean. 



Grid current =5mA 




A 6 

Pressure, torr x lO'' 

Figure 4.15. Typical calibration graphs for ionization gauge 

The hot cathode ionization gauge is not widely used in industrial 
vacuum practice because {i) the gauge head is fragile and easily 
damaged, («) the filament is easily burnt out by careless operation, 
{in) the small ion current produced at low pressures necessitates the 
use of high gain amplifiers which increase the expense and complexity 
of the equipment, and {iv) a cold trap must be used. 

The limits of operation of this form of ionization gauge are from 
about 1 " ^ to 10"^ torr. At pressures above 1 " ^ torr the life of the 
filament is greatly reduced ; in addition there is the possibility of a 
glow discharge being formed. The lower pressure limit is set by the 
emission of soft x-rays from the grid, caused by its bombardment by 
electrons. Although the wavelengths of these x-rays are quite long 

94 



1 



IONIZATION GAUGE 

(~100A) they possess sufficient energy to cause the photoemission 
of electrons from the ion collector. Electrically the emission of an 
electron by the collector is equivalent to the capture of a positive ion, 
leading to a total current in excess of that due to the positive ions. 
The photocurrent appears to be independent of pressure and is of the 
same order of magnitude as the ion current at 10"^ torr. 

The photocurrent is greatly reduced in the design due to Bayard 




Figure 4.16. Bayard-Alpert gauge 



Filaments Collector 



and Alpert^i^' which is shown in Figure 4.16. The positions of the 
cathode and the collector are reversed, and the latter consists of a 
central wire of greatly reduced surface area compared with the 
normal design. The grid is again a wire spiral between the collector 
and cathode. With this arrangement the x-radiation falling on the 
collector is greatly reduced because of its smaller area, and pressures 
down to 10 ~^^ torr can be measured. 

It should also be noted that the ionization gauge has a pumping 
action. The positive ions flowing to the collector are effectively 
removed from the gaseous system and the pressure is consequently 
reduced. This type of pump is discussed more fully in Chapter 3. 

4.9.2 The Penning Gauge 

Another form of ionization gauge is the cold cathode ionization 
gauge^'^ illustrated in Figure 4.17. Two cathodes are used in the form 

95 



MEASUREMENT OF LOW PRESSURE 

of parallel plates and midway between them is placed the anode 
consisting of a loop of metal wire whose plane is parallel to that of the 
cathodes. A potential difference of about 2 kV is maintained between 
the anode and the cathodes. In addition, a magnetic field of the order 
of 400 gauss is applied at right-angles to the plane of the electrodes by 
a permanent magnet. 

Consider an electron at a position near one of the cathodes. It is 
accelerated towards the anode by the electric field, but the action of 
the magnetic field causes its path to be in the form of a tight helbc 
about the direction of the magnetic field (see Section 1.10). The 
electron generally passes through the plane of the anode loop until 
its path is reversed by the electric field due to the second cathode. 
The electron continues to oscillate in this manner about the plane of 



Electron 
path 



Cathodes 



Anode 



Figure 4.17. Penning gauge 

the anode loop, thus following a very long path so that the chance of 
striking a gas molecule is high even at low pressures. The kinetic 
energy of the electron gained from the electric field is generally large 
enough to cause ionization should such a collision occur. The 
secondary electrons produced by ionization themselves perform 
similar oscillations and the rate of ionization increases rapidly. 
Eventually, the electrons are captured by the anode and equilibrium 
is reached when the number of electrons produced per second by 
ionization equals the number of those captured per second by the 
anode. 

The positive ions created in this process are captured by the cath- 
odes and hence cause a current to flow in the external circuit. The 
drift energy of the positive ions is not large enough to contribute 
significantly to the ionization process. The ion current is measured 
by a miUiammeter in the cathode circuit, since the current is suffi- 
ciently large to be measured directly without amplification. 

96 




IONIZATION GAUGE 

The number of ions produced per unit time is proportional to the 
molecular density and hence to the pressure in the gauge; thus there 
should be a linear relationship between current and pressure, al- 
though in practice the current increases at a smaller rate than the 
pressure. The calibration of the gauge, however, depends on the 
type of gas in the system because of the different ionization energies 
of difTerent gas molecules. 

The electrical circuit associated with the Penning gauge is very 
simple and is shown in Figure 4.18. The gauge is operated from a 
control unit consisting of a rectified but unstabilized a.c. power supply. 
The voltage applied across the gauge head may be standardized by 
using the miUiammeter as a voltmeter and bringing the pointer to a 
fixed mark on the scale. 

The range of the Penning gauge is from about 10"^ to 10"^ torr, 



Gauge 
head 




Figure 4.18. Simplified Penning gauge control circuit 



the upper limit being set by the onset of a glow discharge (see Section 
4.5) and the lower limit by the smallness of the ion current generated. 

One disadvantage of the Penning gauge is the difficulty of initiating 
the build-up of ionization, especially at low pressures. In order to 
start the ionization process it is necessary to produce a few electrons 
near one of the cathodes. At higher pressures {i.e. of the order of 
10"^ torr) general background radiation in the atmosphere (cosmic 
radiation, etc.) which passes through the gauge head usually produces 
sufficient ionization to provide these electrons. However, at lower 
pressures, the probability of their production by this means becomes 
smaller, and eventually is insufficient to start the process. Several 
methods may be used to overcome this difficulty : 

(a) The gauge may be switched on at (say) 10 ~ ^ torr and allowed to 
remain on throughout the pump-down. 

{b) For low pressure starting a radioactive gamma or beta source 
may be brought near the head, the gamma or beta rays producing 
sufficient initial ionization. 

(c) A small filament may be built into the head through which an 
electric current is passed for a second or so, the thermionically pro- 
duced electrons initiating the ionization. 



7 — I.H.V.T. 



97 



MEASUREMENT OF LOW PRESSURE 

4.9.3 The Alphatron Gauge 

Another method of producing ionization in a gas is by bombardment 
with alpha particles. In the alphatron gauge"*' a weak source, 
sealed in a capsule, suppUes a steady stream of alpha particles. The 
rate of emission of alpha particles is essentially constant for long periods 
if a long half-life material is used. The ion current produced by alpha 
bombardment of the gas is measured between a pair of electrodes 
with a potential difference of about 50 V between them, but the cur- 
rent is very small and needs considerable ampUfication before it can 
operate a meter. The advantage of this gauge is that the power 
supply may be obtained from dry batteries. The disadvantages 
include the necessity for radiation shielding and the need for the high 
gain amplifier. On the other hand, the gauge can be used in the 
range 10-10"^ torr. However, this is similar to that covered by the 
Pirani gauge and the alphatron has never become a widely used 
instrument. 



4.9.4 Modified Ion Gauges 

A number of other forms of ion gauge have been developed for use 
in the region 10 " ^-10 " ^^ torr, where the main concern in design is the 
reduction of the x-ray limitation. 

Lafferty"^' used the magnetron principle*!^' in designing a modi- 
fied hot cathode gauge, a diagram of which is shown in Figure 4.19. 
Electrons are emitted from a heated tungsten filament surrounding 
which is a cylindrical, positively charged electrode. A potential 
difference of about 300 V is maintained between the filament and the 
cylinder, and a magnetic field of 200-300 gauss is applied along the 
axis of the cylinder. 

The electrons, under the action of the crossed electric and magnetic 
fields, move in cycloidal paths and if the magnetic field is high 
enough the electrons do not strike the cylinder. Positive ions pro- 
duced by collisions are accelerated towards a negatively charged ion 
collector placed at the end of the cylinder. In theory no electrons 
should reach the positive electrode and hence no x-rays should be 
produced, but in practice electrons deflected from their normal orbit 
by collisions may penetrate to the cylinder and there is an x-ray 
limit at about 10"^* torr. 

The magnetron principle has also been used with a cold cathode 
gauge and the work of Redhead"'" is important in this field. Figure 
4.20 shows Redhead's magnetron gauge"^'. The gas in the gauge 
becomes ionized in a similar manner to that in the Penning gauge. 
The electrons in the discharge are trapped in the crossed electric and 

98 



IONIZATION GAUGE 

magnetic fields so that bombardment on the anode is very small. 
The magnetic field is not sufficiently large to contain the positive ions 
and these are collected at the cathode. One of the problems associ- 
ated with this gauge is field emission of electrons at the cathode caused 
by the high electric field, and this electron current tends to swamp the 



Magnetic 
field B 




-Ion collector 



— Filament 



— Magnet 
•Anode 

■Shield 



Figure 4.19. Hot cathode magnetron gauge {from Lafferty"^' 
courtesy of The American Institute of Physics) 



by 



ion current. The effect is greatly reduced by the use of auxiliary 
cathodes placed between the anode and the true cathode. Field 
emission now takes place at the auxihary cathodes but this current is 
not measured. This gauge operates down to pressures of about 1 ~ ^^ 
torr. 

An alternative form of this gauge is the inverted magnetron due to 
HoBSON and Redhead"^- ^"^ which is shown in Figure 4.21. The 

99 



MEASUREMENT OF LOW PRESSURE 

Magnetic field B 



Anode 




Cathode 



(/')Ion current 
amplifier 



Figure 4.20. Redhead's magnetron gauge (from Ked- 

head"*', by courtesy of The National Research 

Council of Canada) 



B 



5-10 kV:= 




Anode 



Ion current (0^ 
amplifierKly 



Auxiliary 
cathode 



Figure 4.21. Inverted magnetron gauge {after Hobson 
and Redhead"")) 



100 



MASS SPECTROMETERS 

positions of the anode and catiiode are reversed but the principle of 
operation is the same as in the magnetron gauge. Again, an auxiliary 
cathode is used to reduce the effects of field emission. 

The modulated ion gauge represents another means of reducing the 
x-ray limitation. Redhead*^^^ developed a form of the Bayard- 
Alpert gauge which has a fourth electrode situated between the grid 
and the collector. This electrode is connected in turn to the grid and 
the collector and the difference in ion currents in the two cases may be 
correlated with the pressure. It is assumed that the x-ray effect is the 
same for both readings and is thus cancelled out in taking the differ- 
ence. In this way pressures down to 10 "•'^ torr can be measured. 
Another development along these lines by Sghuhmann'^^', uses a 
screen between the grid and the collector to prevent x-rays falling on 
the collector. 

4.10 Mass Spectrometers 

The mass spectrometer has proved a most useful instrument in the 
measurement of pressures below about 10"* torr, as it enables the 
partial pressures of the various constituents of the gas to be determined. 
Partial pressures as low as 10 "-"^^ torr have been measured. There 
are several forms of the instrument but the general principle of opera- 
tion is common to them all; the gas molecules are firstly ionized, then 
accelerated and finally separated into groups according to their 
masses. 

The means of ionization is fairly standard in all forms of mass 
spectrometer, the gas molecules being bombarded by thermionically 
produced electrons. The positive ions are collected by a negatively 
charged electrode and the resulting current amplified and then 
measured by a sensitive galvanometer. Alternatively, the ions may 
be made to fall on the cathode of a photomultiplier, the high current 
gain of which eliminates the need for an external amplifier. The 
output current may be measured on a milliammeter and the mass 
spectrum scarmed manually. Alternatively, automatic scanning may 
be used and the spectrum displayed on an oscilloscope or a chart 
recorder. 

4.10.1 Magnetic Deflection Mass Spectrometer 

In this form of the instrument the positive ions are accelerated 
through a potential difference W and the n pass into a uniform 
magnetic field B with a velocity V2eWlme. Then, as described in 
Section 1.10, they travel in a circular path whose radius is given by 
r = {llB)V2WmJe. Thus for fixed values of W and B, and a 

101 



MEASUREMENT OF LOW PRESSURE 

common point of origin, ions are separated into groups depending 
on their values of TUeje. The greatest separation of any two groups 
occurs when the ions have traversed a semicircle. 

For vacuum work the instrument is usually constructed with a single 
fixed position electrical collector. With this arrangement and for 
given values of W and B, only ions of a particular value of m^/e 
corresponding to the appropriate value of r fall on the collector. In 
order to detect the presence of other ions, B or W must be changed ; 



Ion 
accelerating 
electrode 

(+ve) 



^Electron 
/ accelerating 
/ electrode 
(+ve) 



Collector 



Ion source 




Limits of 
magnetic field 

Ion paths 

Figure 4.22. Magnetic deflection mass spectrometer 



it is more convenient to keep B fixed and vary W so as to scan the 
whole spectrum of w^/e values. The scale of the meter measuring W 
may be calibrated in mass numbers on the basis of singly charged 
ions. The layout of such an instrument is shown in Figure 4.22. 

It, is shown above that the greatest resolution of the ions is ob- 
tained by using a deflection of 180°. However, this requires a 
relatively large gauge head and, more importantly, a magnet of large 
pole-piece area since the field must be uniform over the whole ion 
path. More compact and lighter spectrometers with deflections 
of 120°, 90° or 60° are described by Neir'^s) and Davis and 
Vanderslice^^*', but the resolution of these instruments is not so 
good. 

102 



MASS SPECTROMETERS 

4.10.2 Trochoidal Mass Spectrometer 

In this form of mass spectrometer*^^' the positive ions are acted 
upon by crossed static electric and magnetic fields. It can be readily 
shown that in this situation the paths of the ions are trochoidal with 
parameters which are functions of me/«, the electric field E and the 
magnetic field B. The instrument illustrated in Figure 4.23 has two 
fixed slits, and by adjusting the electric field, ions of a given value of 
mje which originate behind one sUt can be caused to pass through the 
other and fall on a collector. The mass spectrum can be scanned by 
varying the electric field over wide limits, keeping B constant. 







+ve 




Ion source 

\ 

CI -ILi \ I . 

\ 
\ 
\ 
\ 
■s. 








CoUe 




1/1 1 

/ 

y 

^_ .---■' Ion pathi 




B© 


-ve 


Magnetic 
field 







Figure 4.23. Trochoidal mass spectrometer 



4.10.3 Omegatron 

This instrument, which is illustrated in Figure 4.24, may be regarded 
as a resonance mass spectrometer and was developed by Sommer, 
Thomas and Hipple'^s) and further by Alpert and Buritz«''>. 
The ionized gas molecules are acted upon by crossed electric and 
magnetic fields E and B respectively. B is constant but E is supplied 
by a sinusoidally varying radiofrequency source. 

In the absence of the electric field the ions would move in circular 
paths with an angular velocity coq given by coq = Bejme. Thus kjo is 
inversely proportional to the mass of the ions. The application of the 
electric field having the same frequency cuq causes the ions to move in 
spiral paths in a similar manner to the action of a cyclotron, and for 
ions having a particular mass there is a corresponding resonant 
frequency so that the radius of the spiral increases rapidly. When the 
radius reaches a certain value the ions strike a collecting electrode. 
Ions having different values of mje do not resonate at the same 

103 



MEASUREMENT OF LOW PRESSURE 



frequency and continue to circulate in orbits of small radius at the 
centre of the system. The mass spectrum is thus obtained by varying 
the frequency of the electric field so that resonances at other values of 
OTe/e are obtained. 



Magnetic field 
B 




Electron accelerator 



Filament collector "-.f. signal 
Figure 4.24. Omegatron mass spectrometer [after Alpert and Buritz'^") 



4.10.4 Quadrupole Mass Spectrometer'-^^'' 

This is a form of mass spectrometer which does not require a 
magnetic field and, in consequence is much less bulky than the other 
models. The instnmient is illustrated in Figure 4.25 and consists of 
four cyUndrical rods to which are applied a combination of d.c. and 
a.c. potentials. For a given applied frequency, it can be shown that 
only ions of a particular value of m^je pass through the spectrometer 
to the collector. Ions of other values oim^je are collected by altering 
the a.c. frequency. 

A modification of this instrument is the monopole spectrometer. 
This consists of a single cylindrical rod and two plane electrodes, 
which act as reflectors giving three electrostatic images of the rod, 
thus completing the quadrupole arrangement as before. 

104 



MASS SPECTROMETERS 

Collector 




r.f. signal 
+ d.c. voltage 



Pattn of 
'in phase' 
ions 



r.f. signal 
+ d.c. voltage 



Filament! | | Anode 

Figure 4.25. Quadrupole mass spectrometer 

4.10.5 Mass Spectra 

Each pure gas produces its own characteristic mass spectrum and a 
mixture of gases gives a series of ion current peaks at mass numbers 
corresponding to the various components of the mixture, the relative 
heights of the peaks giving the proportions of each component. Thus 
the partial pressures of the components can be determined if the 
instrument is suitably calibrated. In practice the analysis is compli- 
cated by the following effects: 

(a) The existence of multiply ionized molecules of the gas. For 
example, peaks due to nitrogen could appear 



at mass number 14 due to Ng * , 
at mass number 9^ due to N^ + + . 

The magnitudes of the peaks decrease as the degree of ionization 
increases. 

(Jb) The presence of dissociation products. For example, the 
ionization process can dissociate carbon monoxide into CO"^, C* and 

105 



MEASUREMENT OF LOW PRESSURE 

O* which would give peaks at mass numbers 28, 12 and 16 respec- 
tively. 

(c) The existence of isotopes of the gas. For example, nitrogen 
peaks occur 

at mass number 30 due to Nj^^, 
at mass number 29 due to N"N^^, 
at mass number 28 due to Ng^*. 

((/) The fact that two or more gases may have the same molecular 
weight. For example, a mixture of nitrogen and carbon monoxide 
gives peaks 

at mass number 30 due to Na^^ and C^^O", 

at mass number 29 due to N^N^^, C^O^^ and C^^O", 

at mass number 28 due to Ng" and Ci^O^^ 

(«) The possibility of multiply ionized molecules of one gas appear- 
ing at the same mass number as a singly ionized molecule of another 
gas. For example, in a mixture of methane and oxygen some of the 
peaks in the spectrum may appear 

at mass number 16 due to 0^^^^^, C^ms''^,0^^'-, O^^^^ +, 
at mass number 17 due to C"H4i + , 0" + , O^""-^, 

These situations can be analysed in simple cases from a knowledge 
of the relative abundances of the various isotopes and the probabilities 
of multiple ionization. However, in more complicated instances 
such as those occurring in the analysis of the breakdown products of 
the complex molecules of diffusion pump oils, the analysis is much 
more difficult. Details of such analyses are given by Craig and 
Harden'^*'. 

A further use of the mass spectrometer is in the field of the leak 
detection and this is discussed in Chapter 7. 



4. 1 1 Calibration of Vacuum Gauges 

In the majority of cases the vacuum gauge is used merely as an indi- 
cator of the order of magnitude of the pressure or molecular density 
within the system, and the scale provided by the manufacturer 
suffices for this purpose. There are occasions, however, when it may 
be necessary to check the calibration or to have a more precise know- 
ledge of the pressure. A number of methods of calibration are avail- 
able and a few of the more important ones are now described. 

106 



calibration of vacuum gauges 

4.11.1 McLeod Gauge Method 

Calibration may be effected against a McLeod gauge in the pres- 
sure range 10-10"* torr. The McLeod gauge is an absolute instru- 
ment, independent of the gas used, in which the pressure may be 
calculated from a knowledge of the dimensions of the instrument 
(see p 76) . The gauge must be used carefully, however, in order to 
eliminate the errors previously discussed, but provided adequate 
precautions are taken this method of calibration is accurate to about 
1 per cent within the pressure range given. 

4.11.2 Expansion Method^^'^^ 

A small known volume of gas at atmospheric pressure is allowed to 
expand into an evacuated vessel of large known volume, and by 
applying Boyle's Law the new pressure can be calculated accurately. 
The expansion process may be repeated to produce accurately known 
pressures between 10 ~* torr and 10 ~^ torr, against which gauges can 
be calibrated. The chief source of error in this method is the outgas- 
sing of the walls of the expansion vessels but this can be minimized 
by thoroughly baking the system under vacuum before use. 



4.11.3 Gaseous Flow Method 

Use is made of the flow eqn 2.2, q 



C{pi—p2), from which the 




Figure 4.26. Gauge calibration unit 



pressure difference {px—p2) across a known conductance C can be 
calculated if the throughput q can be measured. This method is 

107 



MEASUREMENT OF LOW PRESSURE 

described by Normand*^^' and the type of equipment used is illu- 
strated in Figure 4.26. The conductance is in the form of an orifice 
or a short tube, and the value of C is calculated from the dimensions. 
The gas inlet is via a narrow bore tube from a large reservoir where the 
pressure is in the range measurable by a simple manometer. The 
throughput is calculated by measuring the rate of decrease in pressure 
of the reservoir gas. The gauge to be calibrated is used to indicate 
j&i. Providing the rate of gas flow is small the pressure pi will change 
only slowly with time and hence during the time required for a pressure 
observation, p^ may be considered constant. If p^ » p2 the true 
value of />i can be calculated from eqn 2.2, since then px = qjC. 
The gauge may be calibrated over a range of pressures either by 
varying the throughput or by using a series of orifices of different 
conductances (see Oatley'^^'). 

4.11.4 Dynamical Method 

This method enables calibration to be made against a McLeod 
gauge but at pressures beyond its normal range. The equipment is 
similar to that shown in Figure 4.26 but with the addition of a fast 
closing valve in the pipeline between the gas reservoir and the 
chamber. The pressure in the chamber is indicated by the gauge to 
be calibrated and also by a McLeod gauge. Let this chamber have 
a volume V. A steady pressure j&i as measured by the McLeod gauge 
is established and then the valve is closed. The pressure in V falls 
exponentially as the gas is piunped away through the conductance 
and it can be shown that the pressure pt after a time t is given by 



pt = pi exp 



(-?) 



Thus the pressure in the chamber after a time t can be accurately 
calculated and compared with the reading of the gauge under 
calibration. For example, if V is 10 1. and C = 1 1. sec~^, the pres- 
sure falls from 10"^ to 10 ~* torr in 70 sec, and thus calibration to 
very low pressures may be readily carried out. 

The main precaution to be taken in this method is to ensure that the 
amount of gas from the walls of the chamber is small compared with 
that pumped from the closed vessel. 

4.12 Interpretation OF Gauge Readings 

Most commercial non-absolute gauges have scales which are cali- 
brated for clean dry air, and for many purposes this calibration is 
sufficiently accurate. However, it must be borne in mind that as the 

108 



references 

pressure in a system falls the gas composition changes. For example, 
in an unbaked system the predominant gas at pressures down to 10"^ 
torr will be air, but below this pressure the proportions of water 
vapour and hydrogen, due to desorption from the walls, will increase 
rapidly. In a baked ultra-high vacuum system the predominant gas 
is most likely to be hydrogen. The composition of the residual gas 
is generally unknown and thus the gauge reading cannot be corrected. 
Consequently for accurate pressure measurement, especially at low 
pressures, partial pressure analysis by means of a mass spectrometer 
must be made. 

The reading of the gauge also depends on its position in the system. 
For example, if a cold trap is employed a gauge will indicate a lower 
pressure on the pimip side than that on the vessel side. This is due 
to the removal of condensable vapours by the trap and also because of 
the pressure drop across the trap impedance. In addition, a gauge 
'looking' at a cold trap will read a lower pressure than one, at the 
same location, but 'looking' into the vessel. This is due to (i) the 
condensable components being frozen at the trap and hence not being 
reflected into the gauge, and (m) the non-condensable gases reflected 
from the cold trap having velocities corresponding to the cold trap 
temperature, and hence causing thermal transpiration from the gauge 
head. 



references 

1 Carr, p. H. Vacuum 14 (1964) 37 

2 McLeod, H. Phil. Mag. 48 (1874) 110 

3 Venema, a. and Bandringa, M. Phillips tech. Rev. 20 (1958) 145 
* IsHii, H. and Nakayama, K. Vacuum Symp. Trans. (1962) 519 

^ RoMANN, M. P. Vide 3 (1948) 522 

« Knudsen, M. Ann. Phys. 31 (1910) 633 

' Leck, J. H. Pressure Measurements in Vacuum Systems, p 135. Chapman and 
Hall, London, 1964 

» PiRANi, M. Verh. dt. phys. Ges. 8 (1906) 686 

^ Varicek, M. Vacuum Symp. Trans. (1962) 483 
1° Dushman, S. and Found, C. G. Phys. Rev. 17 (1921) 7 
^^ RiDENOUR, L. N. and Lampson, C. W. Rev. scient. Instrum. 8 (1937) 162 
" Bayard, R. T. and Alpert, D. Rev. scient. Instrum. 21 (1950) 571 
" Penning, F. M. Physica, Eindhoven 4 (1937) 71 

" Downing, J. R. and Mellen, G. Rev. scient. Instrum. 17 (1946) 218 
1= Lafferty, J. M. J. appl. Phys. 32 (1961) 424 
^^ Parker, P. Electronics, p 50. Arnold, London, 1952 
" Redhead, P. A. Adv. Vacuum Sci. Technol. 1 (1960) 410 
i« Redhead, P. A. Can. J. Phys. 37 (1959) 1260 

" HoBSON, J. P. and Redhead, P. A. Adv. Vacuum Sci. Technol. 1 (1960) 384 
=" HoBsON, J. P. and Redhead, P. A. Can. J. Phys. 36 (1958) 271 
=1 Redhead, P. A. Rev. scient. Instrum. 31 (1960) 343 

22 ScHUHMANN, W. C. Vacuum Symp. Trans. (1963) 428 

23 Neir, A. O. Rev. scient. Instrum. 31 (1960) 1127 

2* Davis, W. D. and Vanderslice, T. A. Vacuum Symp. Trans. (1961) 417 

109 



MEASUREMENT OF LOW PRESSURE 

=5 Bleakney, W. and Hipple, J. A. Phys. Rev. 53 (1938) 521 

=« SoMMER, H., Thomas, H. A. and Hipple, J. A. Phys. Rev. 82 (1951) 697 

" Alpert, D. and Buritz, R. S. J. appl. Phys. 25 (1954) 202 

=« Paul, W., Reinhard, H. P. and von Zehn, V. Z. Phys. 152 (1958) 143 

2^ Craig, R. D., and Harden, E. H. Conf. fundam. Problems low Press. Meas., 

Paper 12. National Physical Laboratory, Teddington, 1964 
3° Fryburg, G. G. and Simons, J. H. Rev. scient. Instrum. 20 (1949) 541 
" NoRMAND, C. E. Vacuum Symp. Trans. (1961) 534 
32 Oatley, C. W. Br. J. appl. Phys. 5 (1954) 358 



110 



THE CONSTRUCTION OF VACUUM SYSTEMS 



5.1 Introduction 

Vacuum systems are constructed of either metal or glass or a combina- 
tion of these two materials, the choice of material depending largely 
on the particular application for which the system is required. Both 
materials are readily shaped into the required forms but metal 
systems can be constructed to closer dimensional tolerances. The 
outgassing constants of the two materials are similar. 

Metal systems are strong mechanically but may be attacked 
chemically by some gases and vapours. In addition, modification of a 
metal system generally means its dismantling and removal to a 
workshop. 

Glass is fragile but is easily cleaned and is generally not attacked by 
the chemicals likely to be encountered in vacuum practice. Modifica- 
tions to glass systems by glassblowing can be carried out in situ. 
Glass tends to be more porous than metals, a fact which is important 
at very low pressures. Glass also has the advantage of transparency. 



5.2 Metal Systems 

A vacuum system may be constructed from any metal or alloy which 
is not porous and which can be suitably machined or fabricated into 
the required form. The metals which have been most frequently 
used for vessels are brass, copper, aluminium, mild steel and stainless 
steel. These materials should be in the rolled, drawn or forged 
condition; castings tend to be porous and should never be used. Of 
these metals, stainless steel is the best from the vacuum point of 
view, because of its resistance to oxidation and corrosion and also 
because it can be heated to high temperatures for degassing without 
decomposing or losing strength. On the other hand, stainless steel is 
expensive and is rather difficult to work; mild steel is often used in 
large plant where it is not required to degas at a high temperature. 
Mild steel can be soldered and welded easily but is prone to rusting, 

111 



CONSTRUCTION OF VACUUM SYSTEMS 

although this defect can be overcome by painting the outside of the 
system and electroplating the inside with cadmiimi. 

The other materials listed above have the merits of cheapness and 
ease of working but are less satisfactory from the vacuum aspect. 
Copper is readily soldered but oxidizes easily and is soft; brass is 
harder but is not recommended for use at elevated temperatures as 
the zinc content tends to distil. Aluminium is difficult to solder, is 
fairly soft and its low melting point limits the temperature at which it 
can be used. 

Valves, baffles, traps, etc., are generally made of mild or stainless 
steel although sometimes copper and brass are used. 

The basic units of the vacuum system are connected together by 
pipeUnes which range in size from \ in. diameter tubing (often used 
for the fore lines of laboratory equipment) to pipes of 36 in. diameter 
used in coupling the largest vapour pumps. Copper or mild steel 
tubing is most frequently used on the fore side of the assembly, and 
copper, mild steel or stainless steel on the high vacuimi side. The 
piping should be seamless to minimize leakage. 

Complete metal systems are made by joining together a number of 
machined components, and this method of construction enables the 
precise location of the components to be effected. The joints between 
the components may be permanent or demountable. In general, a 
demountable joint is more liable to leak than a permanent one and 
consequently as many of the components as possible should be 
permanently joined. 

5.2.1 Permanent Joints 

Permanent joints are made by welding or by hard or soft soldering, 
the greatest care being taken to avoid pinholes or inclusions in the 
fillet of flowing metal. Soft soldered joints, although much easier to 
make than welded or hard soldered ones, are much weaker mechani- 
cally. Welding may be performed by oxyacetylene torch, electric 
arc (sometimes in an atmosphere of argon) or by an electron beam in 
vacuo. Hard solders, which melt in the temperature range 600°- 
700°C usually contain a large proportion of silver with smaller quanti- 
ties of copper, zinc and cadmium. A borax based flux is used to ensure 
thorough wetting of the metal surfaces. Soft solders are generally 
alloys of tin and lead (an alloy of 50 per cent Sn and 50 per cent Pb is 
requently used). They melt at about 200°C and need a resin flux. 
After both kinds of soldering it is most important to remove all traces 
of flux, which would otherwise act as a copious source of gas within 
the system. A much fuller discussion of welding and soldering is 
given by Pirani and Yarwood*^'. 

112 



METAL SYSTEMS 

Figure 5.1(a) shows the method of making a conventional permanent 
joint, in this case between a pipeline and a flange plate. Wherever 
possible the welding or soldering should be performed on the outside 
of the system. 

In welded joints it is important that a continuous second weld 
should not be made on the inside of the system. Such a double weld 
would enclose a reservoir of trapped gas and a small leak on the inner 
weld would be very difficult to locate. If for the purpose of strengthen- 



I 



^Brazing 



(a) 



nzii 



ib) 



Weld 



Weld 



Figure 5.1. Typical flange joints : {a) high 
vacuum, (A) and [c] ultra-high vacuum 



yUT- 



(c) 



ing the joint some double welding is essential, it is advisable to spot 
weld at two or three points leaving sufficient clearance between the 
weld spots to enable pumping between the components to take place. 

For pipes of diameter up to 1 in. there are available a range of 
brass T- and L-shaped couplings, which have a ring of soft solder 
embedded in each branch. In order to make a permanent joint, the 
end of the pipe is cleaned and inserted into one branch of the coupling 
so that when this is heated the solder melts and forms a smooth joint. 

A sharp right-angled joint in pipelines can also be made by cutting 
each pipe at 45° and soldering or welding the two pipes together. 



8 I.H.V.T. 



113 



CONSTRUCTION OF VACUUM SYSTEMS 

For metal ultra-high vacuum work stainless steel must be used and 
the components must be welded together. In order to avoid the 
formation of oxides within the welds it is customary to use argon arc 
welding. Vacuum or electron beam welding is perhaps more satis- 
factory since the possibility of trapped gas is precluded but more 
elaborate equipment is necessary. Degassing of the resulting 
structure by baking is essential and this process tends to set up 
considerable stresses in the welded joints. In order to minimize this 
stress, the two components should be of the same thickness at the weld. 
For example, if a tube is to be welded to a plate, the plate should be 
recessed to leave an annulus of the same thickness as the wall of the 
tube. The actual weld is then made along this annulus. Such a 
weld is illustrated in Figure 5.1(b). 

An alternative method of welding for ultra-high vacuum work is 
shown in Figure 5.1(c), where the base has been machined to give a 
raised collar to which the tube is welded. This method is not used so 
frequently as it is wasteful of expensive stainless steel. 

5.2.2 Demountable Joints 

Demountable joints in metal systems are made by means of gasket 
seals. For medium and high vacua an elastomer gasket compressed 
between flanges is most frequently used. The elastomers in common 
use are artificial rubbers such as Neoprene, silicone rubbers and Viton. 
Viton has a lower outgassing rate than Neoprene, and can also be 
heated to about 200°C as against about 100°G for Neoprene. On the 
other hand Viton is much more expensive than Neoprene. 

The shape of the gasket may be circular or rectangular, the former 
shape being preferred; the cross-section may be circular, square, 
rectangular or L-shaped. The most common form of this type of 
gasket is the O-ring which is circular in shape with a circular cross- 
section. O-rings are available in standard sizes ranging from ring 
diameters of a few millimetres up to about 40 cm, complete lists being 
available in manufacturers' catalogues. 

5.2.2.1 Demountable Flange Joints — ^The O-ring is located in a groove 
machined in one of the flanges and Figure 5.2 illustrates the typical 
cross-sections which are used for these grooves. The plain rect- 
angular groove of Figure 5.2(a) is easy to machine, whereas the 
dovetail groove in Figure 5.2(b) gives positive retention of the ring 
although it is more difficult to machine. Figure 5.2(c) shows the 
trapezoidal groove which is again easy to machine; the small hole 
indicated enables leak testing to be carried out. In all cases the cross- 
sectional area of the groove should be equal to that of the O-ring so 

114 



METAL SYSTEMS 



that when the second plane flange is bolted on, the O-ring is dis- 
torted until it completely fills the groove and makes a vacuum-tight 
seal. Again, the manufacturers give the exact groove dimensions for 
each O-ring. In general the depth is about 75 per cent of the thick- 
ness of the ring. It is essential that the surfaces of the groove and the 
plane flange should be free of scratches which could act as leakage 
paths. 

When using square or rectangular cross-section gaskets the groove 
should have bevelled edges to achieve a distortive effect similar to 
that of the O-ring. 

In static seals it is not necessary to grease the ring, and in fact the 
use of grease is not recommended as a greased ring may be per- 
manentiy distorted on compression. 





(a) 



ib) 




-Pump-out 
hole 



(0 



Figure 5.2. O-ring grooves : [a) rectangular, (i) dovetail, {c) trapezoidal 



5.2.2.2 Demountable Pipe Couplings — Several forms of this coupling are 
available, all of them making use of O-rings. Typical examples of 
these are illustrated in Figure 5.3. 

In Figure 5.3(a) the respective parts of the coupling are soldered 
to the two pipeline ends, and the seal effected by screwing or clamping 
the coupling together so that the O-ring is distorted between one 
plane and one chamfered surface. 

A solderless form of this coupling shown in Figure 5.3(b) consists 
essentially of a tube whose internal diameter is slightly greater than 
the outside diameter of the tubes to be joined. The ends of the 
coupling are chamfered (to accept the O-ring) and threaded; the 
two clamps are screwed home, distorting the O-rings between the 
chamfered surfaces and washers. 

115 



CONSTRUCTION OF VACUUM SYSTEMS 

The principles of the above couplings may be used to effect pipe 
entries to vessels, valves, etc. Such a coupling is illustrated in Figure 
5.3(c). 

Another form of solderless coupling shown in Figure 5.3(d) is similar 
to that of Figure 5.3(b), except that the coupling tube has an internal 
groove machined near each end. The grooves carry O-rings and the 
tubes to be joined are slid into the coupling from either end. 



__WiW<W« 



(a) 



-vwww 




(b) 



Figure 5.3. Demountable 
pipeline couplings 



WWWMW p VWW— 1 1 



(c) 






(rf) 



Elastomer O-rings 



As in the case of O-ring flange seals, the O-rings should not be 
greased; mechanical loading of the coupling, especially in the case of 
the solderless types, should be avoided. If the chamfered surface is 
smooth and clean and the O-ring is clean, joints with leak rates less 
than 10"^ torr 1. sec"^ can be expected. 

5.2.2.3 High Temperature Gasket Seals — If it is necessary to degas the 
system at temperatures higher than 200°C, elastomer seals cannot be 

116 



METAL SYSTEMS 

used. They are replaced by seals using soft metal gaskets. Several 
forms of this type of gasket are in common use, such as flat copper 
washers, copper gaskets of diamond cross-section and gold, aluminium 
or indium wire. Seals using these gaskets are shown in Figure 5.4. 

One of the metal seals most commonly used consists of a ring of J mm 
diameter gold wire placed between two thick flanges and compressed 




(a) 




n 



ib) 





ISI 



~ n 

up 



n 
□ 





{d) 



1 Ua 


(e) 


^^^ 



^^ 



n 



(c) ^ 

Figure 5.4. Selection of ultra-high vacuum gasket seals: (a) soft metal ring seal, (6) step 
seal, {c) diamond section gasket, (d) flat copper washer seal, (e) shear seal 



to about half its original thickness. The ring is made by fusing to- 
gether the ends of the gold wire and then filing down the joint to the 
same diameter as that of the wire. The flanges must be machined and 
polished very carefully to avoid leakage paths by scratches. 

The metal gasket is not retained in a groove as are the elastomer seals. 
The flanges may be completely flat, thus making easy polishing and 
cleaning and the gasket is held in position by a metal shim, indicated 

117 



CONSTRUCTION OF VACUUM SYSTEMS 

in Figure 5.4(a). An alternative form is the step seal oi Figure 5.4(b), 
in which a step is machined in both top and bottom flanges. The 
gasket fits into the corner on the bottom flange and is distorted by the 
projecting corner on the top flange. 

A large number of steel bolts are used to clamp the flanges together 
and these must be tightened down very carefully, preferably using 
feeler gauges to check the gap between the two flanges and hence 
ensuring even compression of the wire. The expansion of the bolts 
during the baking process must also be considered and, if mild steel 
bolts are used on a stainless steel system, spring washers under the 
bolt heads will maintain the compression on the gold wire. 

When the system is dismantled the flanges are difficult to separate 
and a break bolt is usually incorporated in the design. On dis- 
mantling, the metal gasket is often damaged and must be replaced. 

Metals other than gold (e.g. indium and alimiinium) have been used 
for this type of seal but in general they are not as satisfactory as gold. 
Indium melts at a much lower temperature (150°C) and seals using 
this metal must be water cooled during degassing. Alimiinium tends 
to stick to the steel plates and not only is the ring destroyed but the 
plates require repolishing each time the joint is dismantled. 

Another method of making a seal, shown in Figure 5.4(c), is to use a 
ring of soft copper of diamond-shaped cross-section which, when 
compressed between two plates, deforms and fills the irregularities 
on the surfaces of the plates and produces a good vacuum seal. The 
copper ring becomes work hardened during compression and it is 
necessary to anneal it before it is used again. 

A third method of producing seals is by using a flat soft copper 
washer clamped between two plates on each of which there is a 
projecting ridge which bites into the copper, as in Figure 5.4(d). 
Again the washers can only be used once, but they are relatively 
cheap. 

A modification of this seal is the shear seal shown in Figure 5.4(e). 
In this, the flanges are machined with chamfered steps so that when 
they are clamped together the copper washer is sheared by the steps 
and produces a very good seal. 

5.2.3 Devices for Transmitting Motion 

A rotary or sliding motion is frequently required inside a vacuum 
system, and several methods of causing this are available. 

5.2.3.1 Magnetic Rotator — From the vacuum point of view, this is the 
most satisfactory method as no direct coupling through the wall is 
necessary. In its simplest form, the magnetic rotator consists of 

118 



METAL SYSTEMS 

a magnet outside the vessel and a soft iron block attached to the 
movable component inside the vessel. Movement of the magnet 
causes corresponding movement of the iron block, and hence the 
required movement of the attached component. This method is not 
suitable if a mild steel vessel is being used. 

5.2.3.2 Wilson Seal — A direct driving shaft for rotary motion can be 
carried through the vessel wall by a Wilson seal, which is illustrated 
in Figure 5.5. The shaft must be smoothly polished and the vacuum 



Air 



Elastomer 
washers 




Vacuum 



c :> 




Figure 5.5. Wilson seal for rotation 



seal is made by two elastomer washers with internal diameters slightly 
less than the diameter of the shaft. The pressure of the atmosphere 
keeps these washers pressed against the shaft ; in this case the washers 
must be greased to reduce friction. 

5.2.3.3 Metal Bellows — ^The use of metal bellows to transmit a limited 
reciprocating motion is shown in Figure 5.6. Metal bellows may also 
be used to transmit rotation by means of the wobble drive which is 
illustrated in Figure 5.7. Both of these methods are good from the 
vacuum aspect as there are no sliding seals at which leakage may 
occur. However, the bellows are usually soft soldered to the base 
and this is a source of mechanical weakness. 



119 



CONSTRUCTION OF VACUUM SYSTEMS 




Figure 5.6. Bellows sliding seal 




Vacuum 



Figure 5.7. Wobble seal for rotation 



5.2.3.4 Piston Rings — ^A sliding vacuum seal can be made by using an 
O-ring as a piston ring, two such seals being illustrated in Figure 5.8. 
Figure 5.8(a) shows how this seal may be used for short travel of the 
piston and Figure 5.8(b) shows a form suitable for long travel of the 
piston. 

120 



METAL SYSTEMS 



(a) ib) 

Figure 5. 8. Piston ring seals : (a) short travel, (A) long travel 



5.2.4 Valves for Medium and High Vacua 

All vacuum systems contain valves and a wide range of valve types 
is available. 

5.2.4.1 Medium Vacuum Valve — The simplest vacuum valves are those 
for use in fore lines and one of these is illustrated in Figure 5.9. In this 
an elastomer membrane is forced down into contact with a ridge 




Figure 5.9. Medium vacuum valve 

inside the tube. The valve has a good conductance but the large 
area of exposed elastomer leads to poor outgassing properties. When 
this type of valve is closed the leak rate from the atmosphere is less 
than 10"^ torr 1. sec"-'. 



121 



CONSTRUCTION OF VACUUM SYSTEMS 

If the valve is to be used on the high vacuum side, it must be of 
high conductance and the amount of elastomer used in construc- 
tion must be kept to a minimum to reduce the gas load from this 
source. 

5.2.4.2 High Vacuum and Bellows Valves — Forms of high vacuum valve, 




Figure 5.10. High vacuum valve 



Cam drive 



Figure 5.11. Bellows valve 



Metal bellows 



122 



METAL SYSTEMS 

in which the seal is made between an O-ring and a metal plate, are 
shown in Figures 5.10 and 5.11. The valve operating mechanism 
may be a Wilson seal, an O-ring seal or metal bellows. These valves 
must only be used in such a direction that the pressure difference 
across them does not tend to lift the seal. The open conductance 
of this type of valve is fairly low but the leak rate when closed is less 
than 10"'' torr 1. sec"-"-. 

5.2.4.3 Baffle Valve — ^The baffle valve illustrated in Figure 5.12 is 
frequently used in conjunction with a vapour pump, the object being 
to provide valving action and at the same time to interpose a baffle 
in order to reduce vapour back-streaming. The baffling action of this 
valve results in fairly low open conductance. 




Figure 5.12. Baffle valve 

The action of the valve can be seen by reference to the diagram. 
Rotation of the threaded rod by the external handle via a Wilson seal 
causes the nuts to move in opposite directions along the left- and right- 
handed threads. This movement is transmitted to the baffle plate 
which, when lowered completely makes a vacuum-tight seal with an 
O-ring seated on the base of the valve. Baffle valves are manufac- 
tured in sizes to fit commercially available vapour pumps, their 
conductances ranging between 40 and 30,000 1. sec" ^. The leak 
rate for all sizes is about 10"^ torr 1. sec"-'. 



5.2.4.4 Butterfly Valve — ^Another high vacuum valve is the butterfly 
valve shown in Figure 5.13. The seal is made by an O-ring seated in 
the rim of the central disc. When the disc is swung completely 

123 



CONSTRUCTION OF VACUUM SYSTEMS 

open, the butterfly valve has a high conductance but does not act as a 
baffle. Valves of this type are made in sizes up to 6 in. in diameter: 
a 1 in. diameter valve has a conductance of about 121. sec 
6 in. diameter valve has a conductance of about 2,000 1. sec" ^ 
leak rate is less than 5 x 10"'' torr 1. sec"^. 



^ and a 
The 




Figure 5.13. Butterfly valve 

5.2.4.5 Gate Valve — ^The gate valve shown in Figure 5.14 consists of a 
metal plate which slides across the valve aperture and is vacuum 




Figure 5.14. Gate valve [from Wahl 

et al.^'\ by courtesy of The American 

Institute of Physics) 



sealed by O-rings in the valve casing. The open conductance of this 
valve is high but it provides no baffling action. Gate valves of 

124 



METAL SYSTEMS 

diameters between 4 and 12 in. are manufactured, the conductance 
range being 1,300-20,000 1. sec~^. The leak rates of these valves 
are about 10 ~^ torr 1. sec~^. 

All the three high vacuum valves described above are made either 
with manual or with pneumatic operation. 

5.2.4.6 Magnetically Operated Valves — ^A magnetically operated cut-off 
valve is illustrated in Figure 5.15. When a current passes through the 
coil the magnetic plunger is held up, but if the current is cut off the 
plunger is driven down by the spring and an O-ring seal is made. 



Solenoids 




Figure 5.15. Magnetic valve 



A magnetically operated air admittance valve works on a similar 
principle to that described above, but in this case the failure of the 
current causes the valve to open and admit air to the system. These 
two magnetic valves are frequently housed in one unit and used in 
conjunction with rotary oil pumps. If the current fails the system is 
then isolated by one valve and the pump let up to atmospheric pres- 
sure by the other, so that oil suck-back does not occur. 

5.2.4.7 Needle Valve — ^The needle valve, whose construction is shown 
in Figure 5. 16, is useful for fine control of gas flow. A hardened steel 
needle tapered about \ in. ft.~^ fits inside an accurately machined 
tapered hole so that, when the needle is pressed home, a vacuum seal 

125 



CONSTRUCTION OF VACUUM SYSTEMS 



is produced. As the needle is withdrawn a conducting annulus is 
produced, the area of the annulus increasing with the movement of 
the needle. 




Figure 5.16. Needle valve 



5.2.5 Valves for Ultra-high Vacuum 

The tendency in ultra-high vacuum systems is to avoid the use of 
valves wherever possible. Certainly the conventional forms of 




Gold wire seal 

Soft metal cone 
Hard metal edge 



Figure 5.17. Ultra-high vacuum valve 

vacuum valves involving elastomer diaphragms and O-rings are not 
used because of the high vapour pressure associated with the elas- 
tomers and their inability to withstand baking. 

126 



METAL SYSTEMS 



When a valve is essential, one constructed entirely of metal is 
employed; Figure 5.17 shows a design frequently used. The seal is 
made between a hard metal plunger and a soft copper base. Rotation 
of the vital components is avoided so that the seal always matches up 
at the same place each time the tap is operated. During baking the 
tap must be open otherwise thermal expansion could distort the 
matching pieces and produce a permanent leak. The closing of these 
valves must be carried out carefully and manufacturers usually 
specify the maximum torque which can be applied with safety. 

An ultra-high vacuum gate valve designed by Sheffield*^' is illu- 
strated in Figure 5.18. The seal is made by a copper disc pressing 



Bellows 




Copper disc ^P^^S 
Knife edges 

Figure 5.18. Ultra-high vacuum gate valve {after Sheffield'^') 



against a knife edge machined on the entrance aperture of the valve. 
The disc is mounted on a stainless steel backing plate, which in turn is 
loosely fastened to a stainless steel wedge by a screw which is free to 
move in a slot in the wedge. The wedge slides in a tapered housing 
rigidly fastened to the base of the valve. The wedge is operated by a 
push rod through a bellows mechanism, and a spring ensures that the 
copper disc locates opposite the knife edge before the wedge becomes 
tight. When the wedge is driven into the tapered housing, the disc is 
forced against the knife edge. 



5.2.6 Pressure Sensitive Devices 

Pressure sensitive switches are manufactured for the semi-automatic 
operation of vacuum systems. They are available in ranges [i) 20- 
0-01 torr for use in fore lines, and {ii) 5 x lO'^-lO"^ torr for use in 
high vacuum lines. These switches can be used to control the electro- 
magnetically operated valves described in Section 5.2.4.6. 

127 



CONSTRUCTION OF VACUUM SYSTEMS 

5.2.7 Cold Traps 

A cold trap is frequently used in vacuum systems and is essential 
when using mercury vapour pumps and in the production of ultra- 
high vacua by vapour pumps. The main function of these traps is to 
remove condensable vapours from the system, especially those produced 
by back-streaming from the pump. The essential features of a cold 
trap are high conductance combined with high condensation effi- 
ciency. Thus the vapour molecules attempting to pass through the 
trap should have to make several collisions with the cold walls; in this 
respect the trap should be optically designed, i.e. it should be impos- 
sible for a molecule travelUng in a straight line to pass through the 
trap without at least one collision with a cold surface. The more 



System 

Dewar . 
flaskV 



Vessel 




Figure 5.19. Cold traps 



collisions the molecules make, the lower is the conductance of the trap 
for non-condensable gases; hence in the design of the trap, a balance 
has to be struck between these two considerations. 

Figure 5.19 shows a number of typical cold trap designs. The traps 
are filled with either liquid gas, nitrogen being most commonly used, 
or a mixture of solid carbon dioxide and acetone. The former is 
preferable as a temperature of — 196°C is reached, whereas with 
soUd carbon dioxide and acetone the temperature is — 78°G. 

5.2.8 Oil Creep 

A problem which arises with oil vapour pumps is the creeping of oil 
up the walls of the pump, along the least cool wall of the cold trap, 
and into the working vessel where it produces contamination. This 

128 



GLASS SYSTEMS 

oil creep can be eliminated by designing the cold trap so that a ring of 
the outer wall is cooled to Uquid nitrogen temperature causing the oil 
to be frozen as it attempts to cross this ring. The difficulty in cooling 
the outer wall of the trap is that thermal insulation is poor resulting in 
loss of coolant. A design which avoids this difficulty is shown in 
Figure 5.19(c). 

5.2.9 Cold Baffles 

A high conductance cold trap in the form of a chevron baffle is 
illustrated in Figure 5.20. The baffle may be cooled by water or by 
means of a refrigerator unit working at a temperature of about — 20°C. 
Refrigeration may be carried out by a conventional 'Electrolux' type 
unit or by a thermoelectric cooling element. The cold baffle is 
particularly useful with vapour pimips in that it condenses but does 




Figure 5.20. Chevron baffle 



not freeze the vapour, thus allowing the condensed liquid to run 
back into the pump. On ultra-high vacuum systems where it is the 
practice to use both a chevron baffle and a liquid gas trap, the chevron 
baffle should be placed immediately above the pump. The main 
condensation of vapour takes place at the baffle, the fluid returning to 
the pump, and thus the loss of fluid by freezing at the cold trap is very 
small. 

5.3. Glass Systems 

The second important material which is extensively used in vacuum 
systems is glass. Vessels, vapour and ion pumps, traps, taps and 
pipelines may all be made in glass and even the fore pump could be a 
sorption pump of glass construction. 



5.3.1 Permanent Glass Joints 

The permanent joining of glass components to form a complete 
system is carried out by glassblowing. Such joints, if properly made. 



9 — I.H.V.T. 



129 



CONSTRUCTION OF VACUUM SYSTEMS 

should have the same vacuum properties as the glass walls of the 
system, and as no flux is used in glassblowing there are no cleaning 
problems. Glass joints must be fully annealed to remove stresses 
which would otherwise lead to cracking. If the components to be 
joined are made of glasses of widely different expansion coefficients a 
graded seal must be used. In this type of seal, several grades of glass 
whose expansion coefficients differ only slightly from one another are 
fused together in series, and thus the thermal stresses are minimized. 

5.3.2 Demountable Glass Joints 

Demountability is much more difficult to achieve in a glass system 
than in a metal one. One method is by means of cone joints which 
are carefully ground together and which can be obtained in standard 
sizes up to a diameter of 2 in. The cones must be lubricated and 
sealed with a low vapour pressure grease to preclude degassing by 
heating. 

Demountable joints for glass pipelines can also be made with the 
type of couplings shown in Figure 5.3{d). 

5.3.3 Valves 

Valves for all-glass systems are in the form of stopcocks, carefully 
ground-in and greased. A number of possible designs are shown in 
Figure 5.21. The best valve, Figure 5.2 l{c), is that in which the centre 




(a) (6) 

Figure 5.21. Glass stopcocks 



(c) 



cone is forced home by the pressure of the atmosphere ; in consequence, 
the tap can only be used with the low pressure region below the cone. 
Glass taps are usually only of small bore (up to about 5 cm) and 
thus are of small conductance. Baking is precluded by the use of 
grease and the possible cracking of the tap by thermal expansion. 

130 



GLASS-METAL SYSTEMS 



5.3.4 Mercury Cut-off 

A device which can be used as a valve in glass systems is the mercury 
cut-off, one form of which is illustrated in Figure 5.22. If this is 
used on the high vacuum side, a cold trap must be incorporated to 
freeze out the mercury vapour. 




Level of 
mercury in 
closed position 



Figure 5. 22. Mercury cut-off valve 

5.3.5 Cold Traps and Baffles 

The designs of glass cold traps are similar to those for the metal 
systems, and the traps shown in Figure 5.19 could be adapted for 
manufacture in glass. 

Hollow glass baffles using Uquid coolants can also be used for the 
same purpose as their metal counterparts, although the low thermal 
conductivity of glass compared with metal makes them less efficient. 



5.4 Glass-Metal Systems 

Many vacuum systems are constructed from a combination of glass and 
metal components. 

A common form of this type of system is a glass bell jar mounted on a 
metal plate. The base of the bell jar is ground flat and the vacuum 
seal is made by either an O-ring recessed into the metal plate or an 

131 



CONSTRUCTION OF VACUUM SYSTEMS 

L-section elastomer gasket fitting round the base of the bell jar; in 
both cases the compression of the gaskets is produced by the pressure 
of the atmosphere on the jar. This technique is used for pressures 
down to 10-'= torr, and is claimed to be effective to lO'^ torr although 
degassing by heating can only be carried out at temperatures below 
150°C. 

Another method of joining glass and metal is by means of glass- 
metal seals. The metals used in these seals are alloys with expansion 
coefficients similar to that of glass, and the seal between the alloy 
and the glass is made by glassblowing. 

For soft glasses, alloys of either 50 per cent iron-50 per cent nickel, 
or 74 per cent iron-26 per cent chromiimi are used, whilst the 
alloys for hard glasses are made from iron, nickel and cobalt, typical 
ones being Fernico I (54 per cent Fe, 28 per cent Ni, 18 per cent Co) 
and Kovar (54 per cent Fe, 29 per cent Ni, 17 per cent Co). 

Glass-metal seals can be purchased in various sizes up to several 
inches in diameter. The glass end of the seal is fused to the glass 
component in the system by glassblowing, and the metal end is sol- 
dered or welded into a flange or coupling so that an O-ring joint can 
be made to the metal part of the system. 

Glass-metal seals enable metal ultra-high vacuum taps to be 
incorporated in glass systerhs. 

Demountable joints between glass and metal pipelines can be made 
using the types of coupUng illustrated in Figure 5.3{c) and {d). 



5.5 Electrodes 

It is often necessary to have electrical supplies within a vacuimi 
system. The electrical leads must be insulated and must also make 
good vacuum seals. 

In glass systems, electrical leads may consist simply of tungsten or 
molybdenum wires sealed through the glass walls, both these metals 
having expansion coefficients similar to that of glass. It is important, 
however, to choose a diameter of wire which will carry the required 
electric current without becoming overheated and then possibly 
cracking the seal. 

In metal systems, the electrical lead must be insulated from the 
vessel wall. One means of doing this is by a glass-metal or ceramic- 
metal seal, a typical example of which is shown in Figure 5.23(a) . The 
central lead is sealed into a glass or ceramic bead and a flange of 
Kovar or Fernico sealed round the centre of the bead. The flange 
can then be soldered or welded into a hole in the wall of the system. 

Multiple leads can be made by sealing a number of tungsten wires 

132 



REFERENCES 



into a large glass disc and then sealing this into a Kovar tube, which in 
turn is soldered or welded into the system. 



Electrode. 



Metal ring 

glass 




Elastomer 
0-rings 



Electrode 



lass 
tube 




(a) (6) 

Figure 5.23. Electric lead- throughs for vacuum systems 

A demountable electrode is shown in Figure 5.23 {b). This consists 
of a metal rod passing through a glass or ceramic sleeve and vacuum 
sealed by O-rings. 

REFERENCES 

' PiRANi, M. and Yarwood, J. Principles of Vacuum Engineering, Chap 4. Chap- 
man and Hall, London, 1961 

^ Wahl, J. S., Forbes, S. G., Nyer, W. E. and Little, R. N. Rev. scient. Instrum. 
23 (1952) 379 

^ Sheffield, J. C. Rev. scient. Instrum. 36 (1965) 1269 



133 



THE OPERATION AND DESIGN OF VACUUM 

SYSTEMS 



6.1 Introduction 

A COMPLETE vacuum system is constructed from the various com- 
ponents such as pumps, gauges, pipelines, etc., discussed in detail in the 
preceding chapters. In the present chapter the basic principles 
employed in the design, construction and operation of vacuum 
systems are discussed. 




Figure 6.1. Basic pumping assembly 



Any vacuum system consists essentially of three groups, namely (z) 
the working vessel, {ii) the main pumps, and {in) the fore pumps. 
These groups are linked together by pipelines and subsidiary com- 
ponents such as valves and cold traps. 

The problem facing the designer of a vacuum system is to combine 

134 



INTRODUCTION 

the three groups in the most efficient and economical manner. In 
order to do this he must first determine the characteristics of the 
system in terms of throughput, and must then establish the types and 
characteristics of the pumps required to produce the working pres- 
sure, taking into account the impedances of the pumping lines and 
other components. The final consideration is the time required to 




Figure 6.2. Typical vacuum pumping unit [by courtesy o/Genevac Ltd.) 



produce the desired pressure when the economic operadon of the 
system must be taken into account. 

Figure 6.1 illustrates in block form the basic pumping assembly. 
Pumping is carried out via the main pump and fore line which is 
controlled by the valve Xp, except during the roughing process 
when the roughing line controlled by Xj^ is used. In order to isolate 
the main pump during the roughing process a further valve Xj is 
required between the vapour pump and the working vessei; this valve 
is frequently of the baffle type. An air admittance valve Z^ enables 

135 



THE OPERATION AND DESIGN OF VACUUM SYSTEMS 

the vessel to be brought up to atmospheric pressure and if X, and Z„ 
are closed this can be done without switching off" the main pumps. 
The pressure in the working vessel is indicated by gauge Gi and the 
fore pressure by gauge Gj. 

Figure 6.2 shows a view of a system similar to that described above, 
and the various components can be readily identified. 



6.2 Working Vessel 

The working vessel consists of a glass or metal envelope in which the 
vacuum process is to take place. The nature of the process deter- 
mines the working pressure py, which must be produced inside the 
vessel. As a guide, typical values of /»„, for some common processes 
are given in Table 6.1. 

Table 6.1. Approximate Working Pressure p^ for Some Processes 



Process 


Approximate working pressure (torr) 


Vacuum degassing 




5-100 


Tungsten filament lamp making 
Vacuum arc melting 
Vacuum induction melting 
Aluminizing of mirrors 
Electron beam melting 
Vacuum welding 


10 


10-1 
10-2 
10-3 
10-* 
-4_io-6 
10-* 


Radio valve manufacture 


10 


-s_io-6 


Travelling wave tube manufacture 




10-^ 


Space simulation 


10 


-9_10-16 



The working vessel and its contents are the chief source of the gas 
load of the system. The three sources of gas are (t) true leaks, {it) 
virtual leaks due to the outgassing of the vessel walls, and (m) gas 
released from the working materials as a result of the process being 
carried out. 

It is possible by careM design, construction and testing to reduce 
the gas load from true leaks to negligible proportions and this source 
of gas need not be considered further in this chapter. 

The throughput due to outgassing can be calculated as j = '^K^A^, 
where X, is the outgassing constant of the fth material and A^ is its 
exposed surface area. 

In this context, the value of ^j is the value which exists under 
working conditions when most of the volume gas has been removed 
from the system, i.e. when Vdpjdt is negligibly small. The outgassing 
throughput so calculated will not therefore apply during pump-down 

136 



CHOICE OF PUMP GROUPS 

from atmospheric pressure. The appropriate value of X, can be 
calculated from the data in Table 1.2, and it is reconunended that for 
design purposes the value of /fj at f = 0-1 h be used. 

In good design the throughput must be made as small as practicable 
and this can be achieved in two ways : 

(fl) The materials of construction should be chosen to have the 
lowest value of outgassing constant consistent with economy. It is 
possible to reduce K by degassing the system with a suitable baking 
procedure and by polishing the surfaces. It should be borne in mind 
that the values of the outgassing constants quoted in the literature 
and in Chapter 1 have in general been obtained under rather special 
conditions which may not apply precisely in practical cases ;;|_for this 
reason they should be taken as orders of magnitude only. 

{b) The throughput can be minimized by keeping the surface areas 
as small as possible. Thus the vessel should be as small as practicable 
for the purpose required and all materials not absolutely essential to 
the process should be excluded. It is also important to avoid 
contamination of the vessel and its contents by careless handling. 

The throughput due to gas released during the process is more 
difficult to estimate, since it depends on the actual process and also 
on the previous history of the materials involved. For instance, 
if the operating temperatures of the equipment and the vessel are 
raised in the process, the outgassing rates are increased in a manner 
which is difficult to predict and the calculation of the gas load from 
this source becomes extremely difficult. In these circumstances 
estimates can be made either by reference to similar plant and pro- 
cesses already in existence, or by performing a pilot experiment. 



6.3 Choice of Pump Groups 

The function of the main pimips is to produce the working pressure 
in the vessel, whilst the fore pumps are required for producing and 
maintaining the conditions necessary for the efficient operation of the 
main pumps. The main pimips may be any one, or a combination, 
of those described in Chapter 3, capable of producing pressures less 
than 10 "2 torr; similarly the fore pumps may be any of the ptmips of 
Chapter 3 capable of producing pressures down to 10"'^ torr. 

It is important to choose the most suitable pumps for the particular 
process to be carried out. Possible combinations of main and fore 
pumps are summarized below in terms of the pressure range over 
which they are most efficient and the throughput which they can 
handle. It should be noted that, over a given pressure range, the 
throughput handling capacity can always be increased by using 

137 



THE OPERATION AND DESIGN OF VACUUM SYSTEMS 

several pumps in parallel. Typical examples of the processes for 
which these pump combinations are appropriate are also included. 



6.4 Pump Combinations 

6.4.1 Pressures down to 10'^ torr 

This pressure range can be adequately covered by rotary mechanical 
pumps, preferably with gas ballast. The largest of these pumps are 
capable of handling throughputs up to 3 torr 1. sec~^ at 10"^ torr. 
One of their applications is in the production of tungsten filament 
lamps (see Chapter 8). 

6.4.2 Pressures down to 10 ~^ torr 

As the working pressure approaches 10 "'^ torr both the speed and 
the throughput capacity of rotary pumps decrease and it becomes 
necessary to use an additional pump, such as a Roots or a vapour 
booster. 

For throughputs up to about 10 torr 1. sec" ■'^, a suitable combination 
is a Roots pump backed by a rotary pump. The high critical backing 
pressure of the Roots pump (~10 torr) enables the rotary pump to 
work at pressures and speeds where it can handle these large through- 
puts. This combination has been used in vacuum drying. 

When the throughput is in the range 10-100 torr 1. sec"-^, a vapour 
booster pump backed by a rotary pump can be used. For even 
higher throughputs it may be necessary to use a vapour booster pump 
backed by a Roots which is in turn backed by a rotary pump. A 
typical use of this combination is in the vacuum melting of metals. 

6.4.3 High Vacua in the Range 10~^-10~^ torr 

The standard pumping equipment over this pressure range is a 
vapour diffusion pump backed by a rotary pump. This combination 
can be used for throughputs up to 5 torr 1. sec~^. It finds many ap- 
plications in general laboratory and research work and also in indus- 
trial processes such as vacuum evaporation and electronic valve 
production. 

An alternative arrangement covering the same throughput range is 
a vapour pump backed by a sorption pump. 

When using vapour pumps the choice of fluid must be made between 
oil and mercury. Both are equally satisfactory for vacuum produc- 
tion and throughput handling and the final decision rests with the 
nature of the contamination they cause in the vessel due to the 
inevitable back-streaming. Mercury readily forms amalgams with 

138 



Numerical desigM 

most metals, but if these metals are heated in the production process 
the mercury is re-evaporated without damage to the surface. When 
oil is used (t) it condenses on the free surfaces and is difficult to 
remove, and {ii) it reacts with metals at high temperatures causing 
cracking and the formation of carbonaceous deposits. These effects 
preclude the use of oil in situations where surface properties are 
sensitive to contamination. 

It should be noted that oil vapour from rotary pumps may diffuse 
back into the vessel and cause similar effects, even when a mercury 
vapour pump is being used. 

Another combination in this pressure range is a getter- or 
evapor-ion pump backed by a sorption pump; throughputs up to 
4x 10"^ torr 1. sec~^ can be handled. The great advantage of this 
combination is the absence of pump fluids which means that the 
contamination due to condensation of oil and mercury vapours dis- 
cussed above cannot arise. In addition, because of the use of a 
sorption pump, the system is free from vibration. This combination 
can also be used in general laboratory and research work and in the 
industrial processes mentioned at the beginning of this section. 

6.4.4 Ultra-high Vacua in the Range 10' ^-10'^^ torr 

The vapour diffusion pump backed by a rotary pump can be used 
in this pressure range, although it is necessary to make some modifica- 
tions to the basic system. Thus a cold trap is essential to eliminate 
back-streaming of the pump fluid vapour and in addition it is neces- 
sary to degas the vessel and the main pump by baking. Hence it is 
common practice to employ a second main pump in series to take over 
the pumping action during the degassing process. 

The getter-ion-sorption pump combination can also be used in 
this pressure region. Again, it is necessary to outgas the vessel and 
the pump by baking, but a cold trap is not necessary. 

The applications of these two pump combinations include research 
on surface phenomena and the production of very pure thin films. 

Where high throughputs are involved a cryogenic pump backed by 
sorption or mechanical pumps can be used. One advantage is that 
the pump is actually in the vessel. A particular application of this 
combination is in the simulation of outer space conditions. 



6.5 Numerical Design 

The preceding sections indicate how numerical estimates can be made 
of the working pressure /»„ and the gas load q, and how the most 

139 



THE OPERATION AND DESIGN OF VACUUM SYSTEMS 

suitable pumping combination for the particular process may be 
chosen. 

6.5.1 Main Pumps 

The next stage in the design procedure is to establish the speed 
of the main pumps required to handle this throughput and produce 
the required working pressure. The system is generally designed so 
that its ultimate pressure /»„ is about one-tenth of jf)„. This factor is 
chosen partly as a safety factor but mainly so that a reasonably short 
pump-down time to /i„ can be obtained. 

The effective speed of the main pump in the working vessel is 
obtained from the equation 



5.= 



Pu 



(6.1) 



In general, the pump is connected to the vessel via pipelines, valves, 
etc. In order to allow for the conductances of these components it 
may be assumed as a first approximation that the speed required at 
the pump should be 28^ (when a baffled pump is used then the baffled 
speed should be 2S^. Although a wide range of pumps are available 
it is unlikely that any manufacturer will offer a pump having a speed 
exactly equal to that calculated. The pump having the nearest 
speed (either higher or lower) should be tentatively selected. Let 
this pump have a maximum speed S^. 

The minimum total conductance Cj. of the traps, valves and pipe- 
lines which may be allowed is calculated from eqn 2.7 as 



Vyr 



'Jm'->B 



(6.2) 



6.5.2 Pipeline Dimensions 

The diameter of the pipeline and components should now be chosen 
to be at least equal to the throat diameter of the pump. 

The conductances of the cold traps and valves may be obtained 
from manufacturers' catalogues. Let their combined conductance be 
C(, (calculated from eqns 2.3 and 2.4). Then the minimum pipeline 
conductance C is given by 



C 






(6.3) 



The maximum allowable length of pipeline may now be calculated 
from the appropriate conductance equation in Chapter 2. For all 
pressures below 1 ~ * torr and pipe diameters up to 1 00 cm, the product 

140 



NUMERICAL DESIGN 

pD is less than 1-5 x 10"^ torr cm so that the gas flow is molecular 
(see Section 2.6.4); for pressures between lO"-"^ and 10~* torr the gas 
flow may be in the transitional region where a higher Conductance 
is obtained for given pipe dimensions. However, by treating the 
flow for all high vacuum lines as molecular and hence calculating the 
maximum length L from eqn 2.1 la to be Z, = 12-1 D^jC^, the actual 
conductance obtained may, under certain conditions, be greater than 
that required by eqn 6.3. This only has the effect of giving a some- 
what higher effective speed than is strictly required. 

6.5.3 Physical Layout 

At this stage the physical layout associated with the proposed plant 
must be considered and it may be such that a greater length of pipe- 
line must be used. In this case either (f ) a greater diameter of pipe 
may be used to keep the same conductance and hence the same S„, or 
(ii) the conductance associated with the new pipe length and the cor- 
responding value ofS^ must be calculated and if necessary a new pump 
chosen. Alternatively, it may be possible to use a shorter pipeline to 
allow the use of a pump of lower speed. 

6.5.4 Pipeline Outgassing 

A further effect which may be of significance in connection with 
high vacuum pipelines is the outgassing of the pipes themselves. The 
effect of this is to increase the gas load on the pumps and hence reduce 
the effective pumping speed in the vessel, thus increasing the ultimate 
pressure. 



Pu 

si 






/! 




■* — 2C— >-y- 2C — 


Po 
Sm 


Pipeline 






Pump 


Vesse 

64/20 10 
Figure 6.3. 


Eject 


of pipeline outgassing on pumping speed 



Consider the system illustrated in Figure 6.3. The vessel is con- 
nected by a pipeline of uniform circular cross-section and conduc- 
tance C, to the pump where the pressure is pQ and the speed 5„. Let 
the effective speed in the system be SJ and the ultimate pressure 
/)„. Let 9i be the gas load originating in the vessel and ^2 that of the 

141 



THE OPERATION AND DESIGN OF VACUUM SYSTEMS 

pipeline; then q^ is equal to -nDLK^ where K^ is the outgassing 
constant of the pipe material. Following Venema and Bandringa"' 
^2 may be assumed to originate at the centre of the pipeline and flow 
through half the pipe length. The following gas flow equations are 
then applicable. 

^mpQ = 9l + ?2 

Pu Po- (.+2C 



from which it may be shown that 



SJ = 



S„C 



iC+S^){l+q^lq,)-S„q,l2q, 



(6.4) 



Using this equation the new effective speed in the vessel can be calcu- 
lated and compared with the value of S^ required from eqn 6.1. 
^e' is always less than S^ but if it is only slightly smaller (up to 25 per 
cent) the pump already chosen is adequate. However, if S^' is 
considerably smaller than S^ either a new pump must be selected 
and the calculation repeated from eqn 6.1, or the physical layout 
reconsidered with a view to decreasing the pipe length and hence qz- 

6.5.5 Fore Pumps 

The fore pump must not only be able to handle the gas load when 
the system is at the working pressure, but also be capable of dealing 
with the maximum throughput of the main pump which will occur 
during pump-down. Reference to any main pimip characteristics 
shows that in general this occurs close to the critical backing pressure 
/»6. Let the maximum main pump throughput be q^. Then 6'j,, the 
effective speed required of the fore pump at p^, is given by 



6-.= 



lb 



(6.5) 



If the impedance in the fore Hne can be ignored the displacement S^ 
of the pump required is then calculated from eqn 3.3 as 



'-4-s) 



where j!)i is the ultimate pressure obtainable by the fore pump (usually 
about 0-02 torr). 

The main source of impedance in the fore lines is usually the pipe 

142 



NUMERICAL DESIGN 



itself. In general, the gaseous flow in these lines is viscous both 
during pump-down and at the critical backing pressure. Since the 

To obtain maximum pipe length, 
multiply Lq by p 



A 56 810 



10' 
8 
6 



lo, ft. torr 
20 30 50 



100 200 300 



10'- 
8 
6 

A 



^ 10^ 






8 
6 



10' 





1 








































X 




s 




























, in. 
















s 




!s^D = ^ 












\ 






\ 




















s 


/ \ 


D = 5in. 

1 i 






\, 




\ 




/ 








\ 






\ 


< 




\D = 6in 






\ 


\ 








\ 


\ 


\ 


s 
























































s 


\ 




\ 








\ 








s 


\ 






s 




s 


\ 








s 




S 






s 


s'"' 


\, 










^D-3ln-"\ 










\, 








\ 




\ 






N 


\ 


s 


\, 






\ 






\ 


S 




\ 


s 




s 


\ 




\ 


\ 






> 


































■^ . n ;« 














S^ 




















s 






\ 


















s 




\ 










\ 


\ 








k 


\ 










\ 


\ 








s 




\ 










\ 


\, 








'-^=Hiny 


\ 








\ 


\ 


\ 


s 






s 




\ 


s 


^, 
















































s 


















' 




S 






^ 






s 










s 














\ 










\ 






\"-' '^'■ 






\ 










\ 




\ 










\ 












\ 


s 




















^^=i in\ 


\ 


\ 







10'' 2 3 A 5678 10^ 2 3 A56 BIO'^ 
io.cm torr"' 

Figure 6.4. Estimation of fore line dimensions [from 

Ward and Bunn'^', by courtesy of Engineering 

Materials and Design Association) 

conductance for viscous flow is proportional top, the fore conductance 
is at its lowest value when working at the critical backing pressure. 
However, provided the pipe conductance at this pressure is made at 

143 



THE OPERATION AND DESIGN OF VACUUM SYSTEMS 

least ten times the required ^6 the effect of the pipelines can be ig- 
nored, and the effective speed then equals S^. In practice backing 
speeds are sufficiently low for this criterion to be achieved with reason- 
able values of length and diameter. The maximum allowable 
length of pipe consistent with this criterion may be obtained from 
Figure 6.4, following Bunn and Ward^^^ 



6.5.6 Pump-down Time 

The time required to pump a system down from atmospheric 
pressure to the working pressure is very important from the industrial 
point of view and cannot entirely be ignored in research applications. 
The total pump-down time consists of (i) the time taken for the fore 
pumps to rough the system down to the critical backing pressure of the 
main pump, and (t'i) the time for the main pump to evacuate to the 
working pressure. 

A general expression for pimip-down time t can be obtained as 
follows. In a system of constant volimie the total throughput q at a. 
given instant is given by the simi of that due to volume gas and that 
due to outgassing, i.e. 



.= -.^.. 



(6.6) 



where V is the total volume above the pump considered and qg is 
the gas load at this instant due to outgassing. In general, the value 
of qg is not known but from the arguments put forward in Section 
1.12.2 it may be inferred that it falls fairly rapidly as pump-down 
proceeds, and approaches 2-^i^i when Vdpldt is very small. 

From eqn 1.5 the throughput is also given as q = S^p, and hence 
the time to pump from a pressure p^ to a pressure p2 is given by 



'=-J?W^ 



(6.7) 



6.5.6.1 Rotary Pumps— The pressure range normally used (760-10-^ 
torr) is such that qg is negligible throughout. Further, the variation 
of rotary pump speed with pressure is given by eqn 3.3, and it is 
shown in Section 6.5.5 that with suitable design of the fore lines the 
effective speed S, can be considered the same as the speed at the 
pump throat S^. Hence, eqn 6.7 leads to the pump-down time by a 
rotary pump as 

144 



THE approach TO VERY LOW PRESSURES 



_V CPi dp 
^bJpi P-P'u 



V . 

= e-ln 



Pl-P'. 



P2-P'u. 



(6.8) 



wliere p'„ is the uldmate pressure of the pump. 

The effect of considering the ultimate pressure is only important if 
either /)i or p2 are less than about lOp^. As these pumps are usually 
used for roughing from atmospheric pressure, /)„ is generally negligible 
compared with pi. 

Examination of eqn 6.8 shows that t increases as p2 is reduced and 
hence, if the rotary pump is being used to back a main pump, the 
most economical operation of the system is obtained if the main pump 
is brought into operation at its critical backing pressure. 

6.5.6.2 Vapour Pumps — The speed characteristics of vapour pumps are 
described in Chapter 3 (eqns 3.4 and 3.5) where it is shown that the 
typical characteristic consists of two regions ; thus the pump-down 
time must be calculated in two stages: 

(a) On the rising portion of the characteristic the expression for the 
pump-down time becomes 



ti 



where C„ is the molecular conductance between the pump and the 
vessel. Thus 

{b) The expression for the pump-down time on the second part of 
the characteristic is similar to that for the rotary pump, with the 
difference that the upper pressure limit is p^. Thus 



Se \.Pw-pu\ J 



,dt 



(6.10) 



VPw-pv. 

Since qg dt as a function oip is not generally known, values oit^ and t^ 
cannot be calculated precisely. However, by setting qg equal to zero, 
estimates of the orders of magnitude of <2 and t^ can be obtained 
which, since qg is positive, are always too small. 

6.6 The Approach to Very Low Pressures 

In the preceding sections the design of systems to achieve and main- 
tain certain pressures is considered on the assumption that the 



10 — I.H.V.T- 



145 



THE OPERATION AND DESIGN OF VACUUM SYSTEMS 

outgassing rates of the material of the vessel and of its contents 
are constant when under working conditions. This assumption is 
reasonably valid for many systems since outgassing rates vary only 
slightly over short times. However, if pumping is prolonged then 
the pressure in the vessel falls significantly and to investigate this fall 
it is necessary to consider the time variation of outgassing rates. 

Under conditions of prolonged pumping Vdpjdt may be considered 
negligibly small and eqn 6.6 may be written as 

q = S,p = J^K,A, (6.11) 

The average overall outgassing rate for all the materials within the 
system can be represented by the single empirical eqn 1.23, i.e. 

where K^ is the outgassing constant after 4 hours of pumping. Thus 
eqn 6. 11 may be written as 

S,p„ = AiK.+K^tj^y) 

where p^ is the pressure after t^ hours and A = ^A^ Therefore 

pH=Pu + {Px-pu)tK' (6.12) 

where p^ = AKJS^ is the limiting pressure set by stationary pro- 
cesses {e.g. permeation, argon limit in sputter-ion pumps, etc.), and 
/>! = AKilSg is the pressure after 1 h of pumping. 

Eqn 6.12 may be used to estimate the time required to pump to 
very low pressures, using published data for K^, K^ and y. 



6.7 Examples of Numerical Design 

Example I 

Suppose a vacuum system is required for the arc melting of metals 
for which the working pressure is to be 1 " ^ torr. The volume of the 
working vessel is 2,2501., the surface area of the vessel is 10^ cm^, 
while that of the equipment is 10* cm^. 

(i) Ultimate Pressure 



Pu^ 



10 
= 10- 



torr 



(ii) Gas Load 

Let the outgassing constants of the vessel wall and of the equipment 
after 0-1 h (see Section 6.2) be5 x 10-8and4x 10-Horrl.sec-^cm-^ 

146 



EXAMPLES OF NUMERICAL DESIGN 

respectively. Then, assuming the true leak rate to be negligible the 
total gas load g^ is given by 

?i = 10Sx5xlO-6+10*x4xlO-* 

= 4-5 torr 1. sec"-' 

(iii) Main Pump Requirements 

The effective speed required at the vessel is 

4-5 
? — 
' ~ 10-3 

= 4,5001. sec- 1 

Thus, the first approximation pump speed = 28,, = 9,000 1. sec--*^. 

In the pressure range used and for the high throughput required a 
vapour booster pump must be used. However, there is no single 
vapour booster pump of speed equal to 9,000 1. sec -^ available, and so 
several pumps in parallel must be used. In addition, it is necessary to 
use isolation valves as vapour booster pumps have long warm-up 
times ( ~ 60 min), and it would be uneconomical to have to shut-down 
the pump each time the vessel were opened for the installation and 
removal of workpieces. Suitable pumps would be Edwards 30B4 
vapour booster pumps which have an inlet diameter of 1 6 in. and a 
speed of 4,5001. sec -^. When the appropriate baffle-isolation 
valve is used the speed is reduced to 3,000 1. sec"^. 

To meet the given requirements three such pumps may be used in 
parallel to give a total baffled speed of 9,000 1. sec"-'. In the design 
of the system each pump may be considered separately as handling a 
gas load of 1 -5 torr 1. sec " -^ with an effective speed at the vessel of 
approximately 1,500 1. sec--"^. 

(iv) Pipeline Dimensions 

The minimum pipeline conductance allowable in series with each 
pump is 

1,500x3,000 

3,000-1,500 

= 3,0001. sec- 1 

A pipeline diameter of 16 in. would be used to match the pump inlet 
diameter and hence the maximum allowable pipeline length L (calcu- 
lated from eqn 2.22) is 

_ 12-1(16 x 2-54)3 



3,000 
= 217 cm 



4x16x2-54 



147 



THE OPERATION AND DESIGN OF VACUUM SYSTEMS 

Unless physical layout conditions dictate otherwise, the high vacuum 
pipeline should be made as short as possible, and by carefully arrang- 
ing the booster pumps close to the vessel, the pipeline length could be 
reduced to about 75 cm. The pipeline would then have a conduc- 
tance of 6,300 1. sec"^ giving an effective pumping speed at the vessel 
of 2,030 1. sec~^. The volume of the pipeline is approximately 100 1. 

(v) Effect of Pipeline Outgassing 

The area of each pipeline is (16 x 2-54) x 75 = 9,600 cm^. Let the 
outgassing constant of the pipeline be 5x 10"^ torr 1. sec~^ cm"^. 
The gas load due to pipeline outgassing is 

q^ = 4-8 x 10"^ torr 1. sec"^ 

Hence, the actual effective speed S^' in the vessel, given by eqn 6.4, is 
2,000 1. sec- ^ 

It is now possible to consider using only two 30B4 booster pumps, 
giving a total effective speed of 4,000 1. sec" ^. The ultimate pressure 
produced by this combination would be 



/-« = 



?i 



= 1-12x10-3 



torr 



This is 12 per cent higher than that originally demanded but is 
tolerable and hence it is proposed to use only the two booster pumps. 

(vi) Fore Pump Requirements 

The critical backing pressure of the 30B4 is 5 torr and the maximimi 
throughput is 500 torr 1. sec"-*^ also at 5 torr. In order to reduce the 
fore pump requirements, it is suggested that only one 30B4 be used 
initially to pump to about 5x IQ-^ torr, when the remaining 30B4 
may be brought into operation. 

The backing speed required by one 30B4 at 5 torr 

_ 500, 
5 



■1. 



= 6,0001. min-i 

Thus, the backing speed required at atmospheric pressure 

_ 6000 
- 1-0-02/5 

= 6,0251. min-i 

assuming an ultimate pressure for the backing pump of 0-02 torr. 

148 



EXAMPLES OF NUMERICAL DESIGN 

A suitable backing pump would be a Kinney GKD 310 with a 
displacement of 7,840 1. min-^; the diameter of its inlet is 6 in. and 
from Figure 6.4 the maximum allowable length of pipe of this diameter 
is about 200 m. This length is much greater than is likely to be 
required in practice. The actual length of fore line used depends on 
the physical layout but in this example it is assumed that a line about 
3 m long is used. In order to back both booster pumps a T-fork is 
required. In addition, it is necessary to fit a bypass pumping Une 
round the booster pump and this is assumed to be 4 m long and 6 in. 
in diameter and hence to have a volume of about 75 1. Both the 
roughing line and the bypass line must contain 6 in. diaphragm 
valves for isolation purposes. 

(vii) Pump-down Times 

Rotary Pump 

The total volume F^ to be rough pumped (assuming the booster 
pumps are evacuated, warmed-up and isolated) is that of the vessel, 
plus that of the two high vacuum pipelines between the baffle valves 
and the vessel, plus that of the roughing line, thus 



Fi = 2,250 + 2x100 + 75 
= 2,525 1. 



Hence, from eqn 6.8 

_ 2,525 x 60 760 
'^ ~ 7,840 5 

= 98 sec 
Vapour Boosters 

For the 30B4 the value of z is -0-81 and p^ is equal to 7 x 10-^ 
torr. The total volume V^ to be evacuated by the single 30B4 is that of 
the vessel plus that of the two high vacuum pipeUnes, thus 

Fa = 2,250+2x100 
= 2,450 1. 
Therefore, from eqn 6.9a 

2,450 r. /7xl0-2\-°"1 2,450, 500 



ti = 



3,000 X (-0-81) 
= 31+3 
= 34 sec 



r, /7xl0-2\-''"l , 2,450, 50C 

[i-(^— ) J+3;ooo^"- 



149 



TH£ OPERATION A^fD DESIGN OF VACUUM SYSTEMS 

ta is made up of two parts, one due to the single 30B4 pumping 
between 7 x 10"^ torr and 5x10"^ torr, and the other due to both 
pumps working between 5x 10"^ torr and 10"^ torr. Hence, by 
eqn 6.10 



2,450 , 5x10-2 
In- 



2x2,000 1x10- 



2,450 7x10-2 
" 2,000 5x10-2 + 

= 0-5+1 

= 1 -5 sec 



Both ^2 and t^ will be larger than the values calculated above for the 
reasons given in Section 6.5.6.2, but the total pump-down time can 
be said to be of the order of 3 min, which is sufficiently small to 
satisfy economic considerations. In fact, the main economic con- 
sideration is that the vapour booster pump boilers require about an 
hour to warm up to operating temperature. The complete system is 
as follows: 

A vessel of volume 2,250 1. pumped by two 30B4 vapour booster 
pumps each via a pipeline 16 in. in diameter and 75 cm long, and 
containing a 16 in baffle-isolation valve. Both pumps are backed 
by a single Kinney GKD 310 rotary piston pump via a pipeline 
of 6 in. diameter. A bypass line must be connected between the 
fore pump and the high vacuum pipeline and this is also of 6 in. 
diameter and controlled by a 6 in. diaphragm valve. A similar 
valve is included in the fore line in order to enable isolation of the 
vapour pumps during rough pumping of the vessel. 

Example II 

Suppose a vacuum system is required for the electron beam melting 
of metals, for which the working pressure is to be 5 x 10-^ torr. The 
volume of the vessel is 2,000 1. and the surface area of its internal 
walls is 6 x 10* cm2. The surface area of the equipment inside the 
vessel is 10^ cm2. The process time will be very long so that roughing 
and pump warm-up times are expected to be negligible in comparison. 

(i) Ultimate Pressure 
The ultimate pressure /)„ should be about 10 per cent of j^^, and thus 
/)„ = 5x 10"^ torr 

(ii) Gas Load 

Let the outgassing constants of the vessel wall and of the equipment 
after O-lh be 10"® and 5 x 10"^ torr 1. sec- ^ cm"2 respectively. Then 

150 



EXAMPLES OF NUMERICAL DESIGN 

the total gas load from outgassing is 

qi = 1-1 X 10-^ torr 1. sec-^ 

It is again assumed that the leak rate is negligible. 

(iii) Main Pump Requirements 
The effective speed required in the vessel is 



Pu 

_ Mx 10-^ 
SxlO-*' 

= 22,0001. sec- 1 

In this system it is not proposed to use either an isolation valve (since 
the vapour pump warm-up time is very much less than the process 
time) or a cold trap, and thus the first approximation pump speed 
= 25e = 44,0001. sec-i. 

A suitable pump is an Edwards F 3605 oil vapour pump having a 
diameter of 36 in. and a speed 5„ of 45,000 1. sec'^ 

(iv) Pipeline Dimensions 
The minimum pipeline conductance allowable is 

45,000 X 22,000 
45,000-22,000 

= 43,0001. sec- 1 

The diameter of the pipeline should be 36 in. and hence the maximum 
pipeline length is 

r_121(36x2;54)3_4,36^2-54) 
^- ipso 3(^bx2 54) 

= 95 cm 

(v) Effect of Pipeline Outgassing 

The area of pipeline is i7(36 x 2-54) x 95 = 28,500 cm2. Let the 
pipeline outgassing constant be 10"^ torr 1. sec-^ cm "2. Thus the 
total gas load from the pipeline is 92 = 2-85 x 1 " 2 torr 1. sec - ^ The 
actual pumping speed at the vessel is given by eqn 6.4 as 

S', = 18,8001. sec- 1 
151 



THE OPERATION AND DESIGN OF VACUUM SYSTEMS 

This would give an iiltimate pressure of 



Pu=' 



?1 



_ MxIQ-^ 
18,800 

= 6x 10~^ torr 

This is 20 per cent greater than that required but is within the 
acceptable tolerance. 

(vi) Fore Pump Requirement 

The maximum throughput of the F 3605 occurs at a pressure of 10 ~ '^ 
torr and is equal to 1 5 torr 1. sec " ^. Thus the speed of the fore pump 
required aX p,,{Q-5 torr) 



11 

0-5 



= 1,8001. min-i 
The speed required at atmospheric pressure 

1,800 



1-0-02/0-5 

= 1,910 l.min-i 

assuming p'^ = 0-02 torr. A suitable pump is an Edwards rotary 
oil pump ISC 3000 which has a displacement of 3,000 1. min"^. 

The pump inlet diameter is 2 in. and using the criterion of Bunn and 
Ward {Figure 6.4), the maximum allowable length of fore line of this 
diameter is 200 cm. The complete system thus consists of: 

The vessel pumped by an F 3605 oil vapour pump via a pipeline 
36 in. in diameter and 180 cm long, and backed by a ISC 3000 
rotary oil pump via a pipeline 2 in. in diameter and not more than 
200 cm long. 

Example III 

This example is given only as an indication of the type of calcula- 
tion which can be carried out for systems in which ultra-high vacuum 
pressures are intended. The theory of outgassing is not sufficiently 
developed to allow precise calculations. 

Suppose a metal system is required to operate in the pressure range 
1 X 10""^ — 5x10"^ torr. In the ultra-high vacuum region it is neces- 

152 



EXAMPLES OF NUMERICAL DESIGN 

sary to use a vessel made of stainless steel which, after pumping for 1 h 
has an outgassing constant A"! of 1 ~ '' torr 1. sec " ^ cm ~ *. Let the in- 
ternal walls of the vessel have an area A of 1,000 cm^ and be 0-5 cm 
thick. The vessel is pumped by a sputter-ion pump with an effective 
speed Se of 100 1. sec"^, backed by a sorption pump. 

Then the pressure in the vessel after pumping for 1 h is K-^AjS, = 
10-«torr. 



(i) Using eqns 1 .24 the outgassing constant Kfi of the stainless steel 
after, say, 10* h pumping can be calculated. Thus, from eqn 1.246 
using Do = 6x 10"^ cm^sec"^ 



/n = 



ttLI 



5-76xl0*Z)„ 
= 2,271 h 



Hence, from eqn 1. 24a, 



and 



291 h 



'■h 

= /:„-|-34x 10-1° torr l.sec-icm-2 

Assuming K^ and K^ to be lO^^^ and 3-5 x 10"^° torr 1. sec^^ cm"^ 
respectively, then the pressure pf^ in the system after 10* h is given by 



Pk = 



3-5x 10-1° X 103 
102 

3-5x10-9 torr 



(ii) Alternatively, by making use of the fact that the experimentally 
determined value of y^ for stainless steel is 1 , the time to pump from 
p^ = 10-^ torr to p,^ = 3-5 X 10-^ torr can be determined from eqn 
6.12 as 



with y = 1 . Now 



,, _ IPizhX 

" " \pu-pJ 



Pu 



= 10-1° torr 
153 



THE OPERATION AND DESIGN OF VACUUM SYSTEMS 

and therefore 



t'K 



291 h 



It can be seen that there is a wide discrepancy between the two 
values of the time. This is because in both cases the assumptions made 
are invalid. 

In (i) it is assumed that the outgassing rate falls initially as <""^'^ but 
in practice the initial rate of fall for stainless steel is more nearly as 
r\ In (ii) it is assumed that y is 1 throughout but in practice y 
decreases with time. Thus, the actual time required for pumping 
probably lies between these extreme values. In any case, the pump- 
ing times predicted by both these theories are prohibitively long. 

(iii) The pumping time to reach the same pressure can be consider- 
ably reduced by baking the system at a high temperature. Dayton's 
theory of the effect of baking on outgassing (Section 1.12.3.4) may 
only he. applied to case (i) above. Thus from eqn 1.29a 



K. = 



^1 



where 



D, _ Hl\__\\ 



10- 



Taking// = 10,000 calmol.-S T^ = 27°C, T, = 427°C, Xi 

torr 1. sec-i cm'^ and K^ = 3-5x10-^° torr 1. sec-^ cm'^, gives 



^ 
D^ 



= 1-36x10* 



and 



L = 6-3 h 



Thus, the pumping time has been reduced to a reasonable value. 
Again, in view of the uncertainty of the theory, this value of the time 
must only be regarded as indicative, but it is of an order of magnitude 
which is in agreement with experimental observations. 



6.8 The Operation of Vacuum Systems 

In order to achieve the efficient and safe working of a vacuum 
system, it is preferable to follow a definite operational routine. From 
this point of view, systems may be divided into categories of (i) those 
in which ultimate pressures greater than 10"^ torr are required and 

154 



the operation of VACUUM SYSTEMS 

which therefore do not require baking, and (ii) those in which 
ultimate pressures below 10 ""^ are required and for which a baking 
sequence is recommended. The majority of present industrial 
systems fall into the first category. 

6.8.1 Systems in Which p^ > 10'^ torr 

The following procedure is designed for the operation of systems 
such as that shown schematically in Figure 6.1, in which the main 
pump group may (or may not) include a cold trap. The procedure 
can be adapted to systems in which main pumps are not used by 
ignoring the operations referring to main pumps, baffle valves and 
roughing. 

6.8.1.1 Start-up Procedure — 

{a) All valves should be closed. 

{b) The fore pumps are started. 

[c) The backing valve is opened to exhaust the main pump, the 
pressure being observed on gauge Gg. 

{d) When gauge G^ indicates the critical backing pressure, the main 
pump is brought into operation (together with any necessary main 
pump cooling supplies). If a baffle valve is not used, the system is 
now being fully pumped. If a baffle valve is used the following 
additional operations are necessary. 

(«) The backing valve is closed and the roughing valve opened to 
exhaust the working vessel, the pressure again being observed on G^- 

(f) The cold traps are then filled. 

(g) When gauge G2 indicates critical backing pressure, the 
roughing valve is closed and the backing and baffle valves opened. 

The vessel is now being fully pumped and if the systems have been 
designed correctly the pressure, as indicated by Gj will fall to that 
required. It should be noted that with a new installation, leak 
testing of the backing line and the main pump can be carried out 
between steps {d) and (e) and of the roughing line and the vessel 
between steps (e) and (/). 

6.8.1.2 Intermediate Procedure— ^\m procedure is adopted if it is 
required to open the working vessel to atmosphere without closing 
down the pumps. It is not possible to carry out this intermediate 
procedure if a baffle valve is not used. 

{a) The baffle valve is closed. The vessel is now isolated from all 
the pumps, and air may be admitted so that the vessel can be opened 
and any necessary operations carried out within. 

{b) After the completion of the operation, the vessel is sealed. 

155 



THE OPERATION AND DESIGN OF VACUUM SYSTEMS 

(c) The backing valve is closed and the roughing valve opened. 

(d) Continue as from step (/) in the start-up procedure. 

6.8.1.3 Close-down Procedure — 

(a) The baffle valve, if fitted, is closed. 

(b) The main pumps are switched off. 

(c) The backing valve is closed, isolating the main pumps under 
vacuum from both the vessel and the fore pumps. 

(d) The fore pump air admittance valve is opened where ap- 
propriate. 

((f) The fore pump is closed down (in the case of a sorption pump, 
it is allowed to warm up) . 

(/) If main pump cooling has been used, this should be maintained 
until the pump has cooled to room temperature. 

Ideally the main pump should remain isolated under vacuum and 
hence no attempt should be made to open the baffle valve until the 
vessel above it has been rough pumped. If it is required to admit air 
to the main pump {e.g. for maintenance) the pump must first be 
allowed to cool to room temperature. 

6.8.2 Systems in Which p^ < 10~^ torr and Which Must be Baked 

An ultra-high vacuum system does not normally have a bypass line 
for rough pumping and thus the intermediate procedure is not pos- 




ik^Baked 
region 



Figure 6.5. Ultra-high vacuum 
pumping set 



r-T?^>^ Fore 
|-l h Valve pumps 



Gauge 



156 



REFERENCES 

sible. In addition, the ultra-high vacuum system probably has one or 
more cold traps and possibly two main pumps. A typical ultra-high 
vacuum system is illustrated in block schematic in Figure 6.5. 

6.8.2.1 Start-up Procedure — 

(a) The fore line valve should be closed. 

(A) The fore pump is started up. 

[c) The fore line valve is opened and the main pumps and the 
vessel are pumped. 

{d) When the pressure has been reduced to the critical backing 
pressure of the first main pump the latter may be brought into 
operation. 

(«) Baking is commenced ; the whole of the working vessel and as 
much as possible of the second main pump should be baked. In the 
case of vapour pumps it is essential to bake the top jet. Baking should 
proceed at 400°-450°C for as long as practicable (6-8 h is a con- 
venient time) . 

(/) Baking is terminated and the system allowed to cool to room 
temperature. 

{g) The cold traps are filled, the lower one first. If a refrigerated 
baffle is used, this should be operated before the cold traps are filled. 

(A) The second main pump is brought into operation. 

[i] The pressure in the working vessel is measured and should 
eventually fall to that for which the system was designed. 

6.8.2.2 Close-down Procedure — 

(a) The fore line valve is closed. 

{b) All main pumps are switched off. 

(c) The fore pumps are switched off and vented. 

Ideally, the vessel should never be opened to air but it may be 
necessary to carry out operations which necessitate letting the vessel 
up to atmospheric pressure. In either case, the whole of the start-up 
procedure must be repeated when the system is next used. If the 
vessel has not been let up to atmospheric pressure, the baking time 
needed to give the same ultimate pressure may be considerably 
reduced. 

REFERENCES 

1 Venema, a. and Bandringa, M. Philips tech. Rev. 20 (1958) 145 

^ Ward, L. and Bunn, J. P. Some Aspects of the Design of High Vacuum Systems. 

Engineering Materials and Design Association, London, 1965 
3 Bunn, J. P. and Ward, L. Engng Mater. Des. 8 (1965) 631 



157 



LEAKS AND LEAK DETECTION 



7.1 Introduction 

The limitation on the ultimate pressure attainable in a vacuum 
system is set by the continuous appearance of gas in the system. The 
general term leak is used to describe this phenomenon and may be 
considered under two headings, namely true and virtual leaks. A 
true leak is due to gas entering the system through a hole, whereas a 
virtual leak is caused by the outgassing of the inner surfaces of the 
system. Both these processes produce the same results, i.e. a rise in 
pressure in a system when it is isolated from the pumps, or a finite 
limit to the pressure attainable in a continuously pumped system. 

Leak rate is defined as the quantity of gas which enters, or appears to 
enter, the vessel in unit time. It is thus defined as a throughput and 
the basic throughput equations apply, i.e. 



Leak rate = V -i^ 
dt 



(7.1) 



where Apjdt is the rate of pressure rise in an isolated system of volume 
V. 

Alternatively, for a continuously pumped system, the rate of re- 
moval of gas by the pumps when the system is at its ultimate pressure 
must equal the rate of gas entry by leaks. Thus, if /)„ is the ultimate 
pressure and S^ the effective speed in the vessel. 



Leak rate = S^p^ 



(7.2) 



The practical significance of a leak lies in either the resulting rate of 
pressure rise, or more usually the ultimate pressure. Hence, it is 
impossible to say whether a given leak rate is large or small without 
referring to either the volume of the vessel or to the effective speed. 
It is considerations such as these upon which a decision as to the 
acceptability of a given leak rate is based, and this is discussed more 
fully in Chapter 5. 

In the case of a true leak, gas is forced through a hole by the 

158 



INTRODUCTION 

difference in pressure between the pressure of the atmosphere and that 
in the system. As long as the system pressure remains small compared 
with atmospheric pressure {e.g. system pressures less than 1 torr) both 
the pressure difference and the mean pressure in the leakage path 
remain essentially constant. Thus, whether the gas flow through the 
hole is viscous or molecular, the rate of inflow of gas and hence the rate 
of pressure rise in an isolated system remains constant. This is 
illustrated in Figure 7.1 {a) where pressure is plotted as a function of 
time measured from the instant of isolation. Since the mean pressure 
is about 380 torr the mean free path in the hole is such that for all but 
the smallest leaks the gas flow through the hole is viscous. 



Isolation 



Time 
(a) 




Isolation 



Time 
(b) 



Figure 7.1. Pressurejtime characteristic for a vessel isolated from the pumps 



In the case of a virtual leak in an isolated system, outgassing of the 
walls causes the pressure to rise and hence the rate of resorption 
increases. Eventually a pressure is reached at which the rates of 
sorption and desorption become equal and subsequently the pressure 
remains constant. This is illustrated in Figure 7.1{b), from which it can 
be seen that the characteristic pressure/time graph for a virtual leak 
shows a steadily decreasing rate of rise of pressure. 

In practice, both real and virtual leaks occur together. The char- 
acteristic of this situation is the summation of those of true and 
virtual leaks, as shown in Figure 7.1{c). The rate of pressure rise is 
initially rapid but falls slowly to reach a steady value, characteristic 

159 



LEAKS AKD LEAK DETECTION 

of the true leak. Such a characteristic is readily obtained for any 
particular system by isolating the working vessel, etc. from the pixmps 
and observing the pressure as a function of time. Estimates of the 
virtual and true leak rates can be obtained by measuring the initial 
and final constant slopes respectively and substituting these values for 
dpjdt in eqn 7. 1 . Virtual leaks can only be reduced by the degassing 
methods discussed in Chapters 1 and 6. True leak rates may be 
reduced to insignificant proportions by finding and sealing the hole. 
The remainder of this chapter is devoted to a discussion of true leak 
detection. 



7.2 Leak Detection 

Leak detection methods may be divided into two broad groups: 
(a) Those in which the system under test is filled with gas to slightly 

over atmospheric pressure and the outflow is detected. 

(A) Those in which a device within a low pressure system detects 

the inflow of an externally applied probe gas. 



7.3 Over Pressure Methods 

7.3.1 Painting with Soap Solution 

With a small excess air pressure in the system, suspected areas are 
painted with soap solution, the leak being indicated by bubbling. 
Leak rates of the order of 10"* torr 1. sec"i can be detected in this 



7.3.2 Total Immersion 

In an extension of the method above, the whole assembly is im- 
mersed in water, leaks being indicated by a stream of bubbles. The 
sensitivity of this method is governed partly by the rate at which gas 
issues from the system and partly by the mechanism of bubble forma- 
tion. For air and water the lower leak rate limit is of the order of 
5 X 10"^ torr 1. sec"^, but by using a gas of low viscosity (to increase 
its rate of outflow) and a liquid of low surface tension (to increase the 
probability of bubble formation) Biram and Burrows^^' showed that 
leaks as small as 10"'' torr 1. sec"^ could be detected. 

In the methods given here and in Section 7.3.1, it is important that 
the system is pressurized before the search liquid is appUed or before 
immersion; this prevents liquid from blocking the holes by capillary 
action and from entering the system. 

160 



low pressure methods 
7.3.3 The 'Sniffer' Technique 

In another over pressure method, often known as the 'sniffer' 
technique, gas issuing from a leak in a pressurized system is sampled by 
being drawn through a flexible tube and into a suitable detector. One 
form of this technique, described by White and Hickey'^', uses a detec- 
tor based upon the fact that platinum, at a temperature of about 900°C, 
emits positive ions even in air. This effect is considerably enhanced 
if the air contains a small proportion of a halide gas or vapour. 

Air is drawn by means of a small fan from the sniffer probe into the 
detector head. The head consists essentially of a platinum filament 
which is heated by an alternating current and which indirectly 
heats a platinum cylinder moimted closely around it. A second 
cylindrical electrode is concentrically mounted around the heater 
assembly and is maintained at a steady potential of some 100 V nega- 
tive with respect to the heated cylinder. If the sniffer probe is 
brought near to a leak in the system pressurized with halide gas the 
emission of positive ions by the heated platinum increases markedly ; 
the ions are accelerated to the negative electrode and cause a current 
to flow in the H.T. circuit. After suitable amplification, the current 
can be displayed on a milliammeter. By using Freon as the halide 
gas, it is claimed'^' that leak rates down to 6 x 10"^ torr 1. sec" ^ can 
be detected. 



7.4 Low Pressure Methods 
7.4.1 Introduction 

In these methods the detector is essentially a pressure gauge 
mounted within a continuously pumped system, a leak being indicatea 
by an apparent change in pressure when the leaking air is replaced by 
a suitable probe gas. Alternatively, the detector may be selective 
in that it only responds to a particular probe gas. Thus, leak detec- 
tion can theoretically be carried out using the normal pressure gauge 
and a suitable probe gas; this principle is in fact used directly for 
detecting large leaks (see, for example, the use of the discharge tube 
in Section 7.4.4.1) but cannot be expected to give high leak rate 
sensitivity. For sensitive leak detection a number of essential 
factors must be taken into account : 

(a) The pressure in the region of the detector head must be stable 
so that small pressure changes due to probe gas can be detected. 

[b) The rate of flow of gas past the detector must be sufficiently slow 
for the detector to be able to respond to a small proportion of probe gas, 
but must be sufficiently fast for the time lag between probe application 
and detector response to be fairly short. 



1 1 — I.H.V.T. 



161 



LEAKS AND LEAK DETECTION 

(c) The detector head must be positioned so that gas from all possible 
leaks flows past it. 

The preferred leak detection system, taking into account the above 
factors, is illustrated in Figure 7.2. Although leaks may be located 
without using such a system the full potentialities of a given detector 
head will not be realized. 

The backing space is to stabilize the pressure near the detector 
head D. Valves X-^ and X2 are throttle valves which control the rates of 
flow of gas into and out of the backing space. The valve Xg 
is a bypass throttle valve which controls the proportion of the total 



Vessel under 
'test 



Roughing line^ 




To mechanical 
fore pump 



Figure 7.2. Schematic representation of preferred leak detection system 



system gas passing through the backing space. When the leak is 
large, pumping via X^ and X2 may not reduce the pressure in the 
backing space sufficiently for it to be within the range of the detector. 
In this case Xg is opened allowing some of the gas to bypass the back- 
ing space, thus reducing the pressure at the detector to the required 
level. Blears and Leck'*' analyse the use of such a system on the 
assumption that the gas flow through the hole is purely viscous. 
They show that if the pressure/) in the backing space is measured by a 
gauge of sensitivity ift such that the deflection of the gauge indicator is 
F = ijip, then the change in gauge deflection AF on completely cover- 
ing the leak with probe gas is given by 



AF 



162 



(7.3) 



LOW PRESSURE METHODS 

where is a constant dependent upon the hole dimensions, S (con- 
trollable by valve X2) is the effective speed in the region of the gauge, 
7] is gaseous viscosity, and suffices g and a refer to probe gas and air 
respectively. 

It is apparent that for maximum AF the detector must have a high 
sensitivity to probe gas, low probe gas viscosity, and a small value of 
Sg. The effect of reducing Sg by closing valve X2 is to increase the 
pressure in the backing space and to increase the time lag between 
probe application and detector response. These effects give practical 
limits to the reduction oiSg. Reducing Sg in this manner also reduces 
Sa which, from eqn 7.3, is undesirable. However, Sg can be made 
much smaller than S^ by using a probe gas of high molecular weight 
M, since Sg is dependent on M"^'^ (see eqns 2.16 and 2.7). 

7.4.2 Probe Gases 

Where a selective type of detector is employed, the probe gases used 
are obviously those to which the detector responds. Such probe 
gases are discussed with the relevant detectors. 

For detectors based on normal pressure gauges the probe require- 
ments were discussed in the previous section. A figure of merit for a 
particular probe gas is the substitution sensitivity factor <f>, defined by 
Blears and Leck<^' as the ratio of the change in pressure indication 
when the leak is covered with probe gas to the mean air pressure before 
probing. Thus, probe gases having large values of ^ are desirable. 
Experimental values of (/> quoted by Blears and Leek are given in 
Table 7.1. 



Table 7.1 Values of Substitution Sensitivity Factors (j> 
{after Blears and Leek'*') 



Probe gas 



Pirani gauge Ion gauge 



Butane ('Calor' gas) 

Ether 

Carbon dioxide 

Carbon tetrachloride 

Hydrogen 

Coal gas 



1 

0-7 

0-3 

0-05 

0-4 

0-25 



10 
5 
1 
1 
-0-4 
0-25 



It can be seen that for an ion gauge, the use of butane will give a 
sensitivity at least 40 times greater than coal gas. It should be noted 
that with the exception of carbon dioxide there is a considerable fire 
risk associated with all the probe gases quoted. 

163 



LEAKS AND LEAK DETECTION 

7.4.3 Minimum Detectable Leak 

This is governed by the inevitable fluctuation A/) in the indicated 
air pressure due to instabilities in pump and detector performance. 
The fluctuation is termed the minimum detectable pressure change, 
and is likely to be of the order of 2 per cent of the mean pressure 
indication. From this it can be shown that the minimum detectable 
leak rate is given by 



"'^min. 



<!> 



Again it is apparent that probe gases having high values of <j> are 
required, combined with small values of S^. Further, if the value of 
l^f for a given detector is known, then ^^^j^ can be calculated for any 
leak detection system. However, it is more usual for the value of 
^Lmin. to be quoted in the literature; this may be misleading since it 
applies only to the particular system and probe gas with which it was 
obtained. 



7.4.4 Detectors 

7.4.4.1 Discharge Tube — As stated in Chapter 4, the nature of the 
gaseous discharge is not particularly sensitive to small pressure 
changes, but its colour is very sensitive to the nature of gas. This is 
particularly marked in the change from the pinkish colour charac- 
teristic of air, to an intense blue characteristic of hydrocarbon 
vapours, or to whitish characteristic of carbon dioxide. Thus, volatile 
hydrocarbon liquids (such as acetone or methylated spirit), or carbon 
dioxide are commonly used as probes. 

The elaborate arrangement oi Figure 7.2 is not usually used with this 
detector. The discharge tube is merely situated in the backing line of 
a vapour pump and is used to detect comparatively large leaks of a 
magnitude such that the system cannot be roughed down to the 
vapour pump critical backing pressure. 

7.4.4.2 Pirani Gauge — The response of the Pirani gauge head to a 
probe gas is caused by the change in thermal conductivity compared 
with that of air. The effect is enhanced by using a gas which also has 
a low viscosity. Since the pressure indication of a Pirani gauge is the 
out-of-balance current of a Wheatstone bridge it is most sensitive to 
pressure changes when the out-of-balance current is zero. For this 
reason the control unit is a modification of that used in pressure 
measurement, in which the balancing arm resistance can be readily 
varied to balance the bridge at the pressure in the backing space. 
A typical bridge circuit is shown in Figure 7.3. Further sensitivity 

164 



LOW PRESSURE METHODS 

can be obtained by using a more sensitive galvanometer than is 
usual for pressure measurement. For maximum sensitivity the head 
must also be protected from ambient temperature changes which 
would increase the value of A/. Such protection can be achieved 
either by adequate thermal insulation of the head or by use of a 
compensator head as described in Section 4.8. 1 . 



Balance control 
Coarse 



Fine 



Meter sensitivity 
control 




Pirani head 



3V d.c. 
Figure 7.3. Typical control circuit for Pirani leak detector 



It can be seen from Table 7.1 that maximum sensitivity is obtained 
by using butane as the probe gas, but in practice hydrogen is most 
commonly used. 

7.4.4.3 Ion Gauges— 'Both, hot and cold cathode gauges may be used 
for leak detection. Their response to a probe gas is principally due 
to the change in ionization potential, combined with greater rate of 
inflow due to low viscosity. However, some probe gases also produce a 
marked change in the electron emission characteristics of the filament. 
For example, Lawton^^' describes the decrease of emission when 
oxygen is used as a probe; use has also been made of the increase in 
emission caused by trichlorethylene. With these methods, however, 
there is some risk of permanent filament poisoning resulting in a change 

165 



LEAKS AND LEAK DETECTION 

in gauge calibration, and they are not recommended if the head is 
normally used for pressure measurement. Butane is the best gas to 
use as a probe, as indicated by Table 7.1. 

7.4.4.4 Palladium Barrier Detector — ^This is a form of hot cathode ion 
gauge which is sensitive only to hydrogen. It depends for its opera- 
tion upon the fact that in the cold state palladium is impervious to all 
gases, but when red hot is highly porous to hydrogen. The construc- 
tion of the detector head is illustrated in Figure 7.4. It is normally 
evacuated to a pressure of the order of 10"'' torr. In operation the 
palladium barrier is heated by electron bombardment, the electrons 
being produced thermally and accelerated towards the positively 



Cathode (-v^-lOOV) 



Ion collector 
C — 150V) 



'Kovar tube 




Heater 



Figure 7.- 



~Glass envelope 
Palladium barrier leak detector head 



Palladium 
barrier anode 
(OV) 



charged palladium. Hydrogen probe gas which enters the system via 
leaks, diffuses through the hot palladium and is ionized by collision 
with the electron stream. The resulting positive ions are collected 
by an electrode which is at a negative potential with respect to the 
cathode This causes a current to flow in the collector circuit. The 
current is amplified and displayed as in the hot cathode ionization 
gauge. 

If water or hydrocarbon vapours reach the hot palladium there is a 
high probability that they will be dissociated to produce hydrogen. 
This will cause an erroneous indication of a leak, or a high background 
current against which small leaks will not be detectable. To avoid 
this effect it is essential to use a liquid nitrogen cold trap between 
the detector head and the vapour pumps. 

At the completion of leak detection the head is operated in a hydro- 
gen-free system, the hydrogen being pumped out of the head until the 

166 



LOW PRESSURE METHODS 

ion current indication is a minimum. The limit to sensitivity is 
set by the background current due to the partial pressure of hydrogen 
in the atmosphere and the evolution of hydrogen from the walls of the 
system. 

7.4.4.5 The Halide Detector — ^This detector operates on the same 
principle as the halide sniffer described in Section 7.3.3, and the 
detector head has a similar construction but is mounted within the 
low pressure system. Freon is used as the probe gas. A particular 
advantage of this detector is that it functions satisfactorily at pressures 
as high as a few torr. 

7.4.4.6 Mass Spectrometer — In principle any of the types of mass 
spectrometer described in Chapter 4 together with any probe gas 
may be used for leak detection, since the device can be adjusted to 
respond only to that gas. However, it is more common to use a 
specially designed head of preset gas selection (usually hydrogen or 
helium) with its own backing space, throttle valves, pumping assembly 
and power supplies, etc. In this form, it is potentially the most 
sensitive leak detector available, but to maintain its sensitivity it is 
essential that it be kept scrupulously clean and protected by a liquid 
N2 vapour trap. It also requires considerable skill in operation and 
is expensive compared with the devices already described. It is 
therefore normally only used for the detection of very small leaks on 
clean components. 

Either hydrogen or helium is used as probe gas partly because of 
their low atmospheric abundances, and partly because it is virtually 

Table 7. 2. Summary of the Properties of Low Pressure Leak Detectors 



Detector head 



Useful pressure 
range 
(torr) 



Order of 

magnitude 

ofAp 

(torr) 



Leak rates 
reasonably 
detectable 

(torr 
1. sec~^) 



Probe 



Discharge tube 

Pirani gauge 

Ionization gauge 

Palladium barrier 
Halide diode 

Mass spectrometer 



10-10- = 

1-10-3 

below 10-2 

below 10 -» 
below 5 

below 10-3 



10- = - 



10- 



10-2 

■10- = 

10-8 

io-» 
■ io-» 

fI0-i° 
I10-" 



10- = 

fio-^ 

1.10-* 

fio-« 

(10-'' 

10-' 
10-' 

fio-' 

110-' 



Alcohol 

Acetone 

Butane 

Hydrogen 

Butane 

Hydrogen 

Hydrogen 
Freon 
Hydrogen 
Helium 



167 



LEAKS AND LEAK DETECTION 

impossible to obtain a pressure indication at masses 2 or 4 with any 
other gas. For both gases, the limit of leak detection is set by the 
background current due to internal sources of the gases. Since far 
more hydrogen than helium appears in the system by outgassing, 
it is preferable to use helium especially since its cost is no longer 
prohibitive. 



REFERENCES 



REFERENCES 



1 BiRAM, J. and Burrows, G. Vacuum 14 (1964) 221 

2 White, W. C. and Hickey, J. Electronics 21 (1948) 100 

' Steokelmacher, W. Schweizer Arch, angew. Wiss. Tech. 27 (1961) 3 
* Blears, J. and Leck, J. H. J. scient. lustrum. Suppl. 1 (1951) 20 
= Lawton, E. B. Rev. scient. Instrum. 11 (1940) 134 



7.5 General Leak Detection Procedure 

Leak detection procedures may be classified as proving or location. 
In the former procedure the object is to determine whether or not a 
leak greater than a stipulated acceptable value exists in a component 
whilst that of the latter is to locate the precise position of a leak once 
its existence, or non-acceptable size, has been established. 

Proving may be readily carried out with any of the methods des- 
cribed appropriate to the ranges of pressure and leak rate under 
consideration. For low pressure methods the usual technique is to 
surround the component under test by an envelope or hood of a 
material such as thin rubber sheet or polythene, which is then 
flooded with the appropriate test gas. 

In the location procedure a number of points must be borne in mind : 

(a) It is advisable to prepare a pressure-rise/time character- 
istic of the system to ensure that the true leak rate is within the 
range of the proposed detection system. In this context it should be 
noted that the initial pressure rise will, in general, be much greater 
than that due to true leaks, and that the total true leak may be due to 
the combined effect of several small leaks. 

{b) The greater the stability of the mean pressure at the detector 
head the greater is the sensitivity of leak detection. In order to 
achieve suitable stability careful use must be made of the throttle 
valves, and fairly long times to reach stability must be accepted. 

(c) When using a probe gas which is less dense than air, probing 
must begin at the top of the vessel under test and gradually moved 
downwards. 

{d) Sufficient response time must be allowed after application of the 
probe gas before proceeding to the next test point. This time can be 
as much as 30 sec. 

(«) When testing very large systems it is often necessary for the 
probe operator to be remote from the detector control unit and 
meter. In these situations it is convenient if the detector response is 
converted into a suitable audio signal. Such adaptations are readily 
available for most of the gas sensitive devices. 

168 



169 



SOME APPLICATIONS OF VACUUM 
TECHNIQUE 

8.1 Introduction 

A BRIEF survey of some of the more important applications of high 
vacuum technology is nov^' presented. It is not intended to be 
complete in its coverage nor exhaustive in individual cases but more 
as an indication of the fields where this technology has been success- 
fully employed and of the particular vacuum problems involved. 
It also serves to illustrate the practical implementation of the prin- 
ciples outlined in the previous chapters of this book. 

8.2 Vacuum Coating 

8.2.1 General 

The deposition of coatings of both metals and non-metals is one 
of the major fields of application of high vacuum technique. The 
industrial applications of this process range from lens blooming and 
the aluminizing of mirrors to the manufacture of electronic circuit 
elements and the production of decorative finishes'^ '^^ There is, in 
addition, a considerable research interest in the production and pro- 
perties of films. The main advantage of vacuum deposition is that 
coatings may be produced which are free from contamination, 
particularly oxidation. 

There are two main methods by which vacuum coatings are 
produced. The first is by the evaporation of the required material 
and the subsequent condensation of the vapour on a suitable sub- 
strate. The second method is by sputtering, a process in which the 
metal to be deposited is bombarded with gaseous ions which dislodge 
clusters of metallic atoms so that they can fall as a film on a substrate. 

8.2.2 Evaporation 

The evaporation of materials in a vacuum requires that firstly they 
should be melted, and then the vapour pressure of the molten phase 

170 



VACUUM coating 

should be at least equal to the pressure in the system. Data on the 
variation of vapour pressure with temperature for many elements 
may be found in Turnbull, Barton and Riviere*^\ At reduced 
pressures it is possible to evaporate materials at considerably lower 
temperatures than their normal boiling points. 

The rate of evaporation iJ^ of a heated substance can be calculated 
from Langmuir's equation**' as 



Re = 5-85 X 10 ^pv /•= g cm ^sec 



'J 



(8.1) 



where M is the molecular weight and pv is the vapour pressure in torr 
at T'°K. The amount of the evaporant deposited per unit area of 
substrate depends on R^ and also on the distance between the source 
and the target ; large distances give uniform deposits but are wasteful 
of evaporant. The angle of incidence on the substrate also affects 
the deposition rate and, in general, should be near normal. 



8.2.2.1 Basic Evaporation Equipment — The basic equipment needed for 
the evaporation process consists of a pumping set connected to an 




Electrodes 






Figure 8.1. 
Evaporation unit 



Baffle valve 

and pumping 

unit 



171 



SOME APPLICATIONS OF VACUUM TECHNIQUE 

easily demountable working chamber which contains the evaporant 
source and the substrate. 

The essential features of evaporation equipment are illustrated in 
Figure 8.1, whilst Figure 8.2 shows the internal structure of a typical 
commercial unit. In the upper half of the photograph may be seen a 
series of circular substrate holders, and at the lower left is a multiple 




Figure 8.2. The internal structure of a typical commercial 
evaporation unit {by courtesy of Edwards High Vacuum Ltd.) 

source holder which may be rotated by an external control via a 
chain drive. The source consists of a container to hold the evaporant, 
and a heater whereby the container is raised to the vaporization 
temperature of the evaporant. 

In order to achieve good adhesion between the coating and the sub- 
strate, the substrate must be thoroughly cleaned both chemically and 
physically before use. Details of such cleaning are given by Pirani 
and Yarwood^^'. 

One form of the working vessel is a glass bell jar resting on a metal 
base plate, the vacuimi seal being effected by an elastomer gasket of 

172 



VACUUM COATING 

L-shaped cross-section. An alternative form is a cylindrical metal 
vessel, one end of which is hinged and which is vacuum sealed by 
elastomer O-rings; it is customary to include glass viewing ports so 
that the process can be observed and controlled visually. 

The pumping equipment for industrial plant operating at pressures 
down to 10"^ torr usually consists of oil vapour pumps backed by 
rotary pumps. For lower pressures, getter-ion pumps backed by 
sorption pumps are frequently used to avoid oil contamination. In 
either case it is necessary to maintain a high effective pumping speed 
within the working vessel so that the gas released by the evaporant 
source can be rapidly removed and hence a low pressure maintained. 
It should be noted that the films deposited on the inside walls of the 
working vessel sorb considerable quantities of gas when the vessel is 
opened to atmosphere. This gas is desorbed during subsequent 
pimip-down and may set a high value on the ultimate pressure. 
Hence the inside of the working vessel must be cleaned frequently. 

The pressure in the working vessel is measured by a Penning gauge 
or a hot cathode ionization gauge. 

8.2.2.2 Evaporant Sources — The mode of evaporation depends on the 
chemical reactivity of the evaporant and also on the form in which the 
evaporant is available. A number of different evaporant sources 



r 



mm 



w 



(a) 



W 




W) 



/ 1^^^^ 1 



(e) 




Figure 8.3. Evaporant sources: [a) tungsten spiral, {b) W-shaped 

tungsten wire, (c) tungsten wire basket, {d) ceramic crucible, (e) and 

(/) molybdenum boats 



are in common use and are suitable for material in the form of wire, 
rod or powder (see Figure 8.3). 

One of the most common evaporant sources is a tungsten wire or 

173 



SOME APPLICATIONS OF VACUUM TECHNIQUE 

molybdenum boat heated directly by an electric current (~30A). 
The tungsten wire is shaped into a loose coil or a W-shape and the 
material, in the form of small lengths of wire is placed on each loop. 
When the filament is heated, the material melts, ideally wets the wire 
and flows along the filament and then evaporates. In some cases the 
evaporant does not wet the wire but forms a series of beads. Evapora- 
tion is then much slower as the heat transfer from the wire to the bead 
is reduced. If the material is available only in powder form, it can be 
evaporated from a boat formed from sheet molybdenum. 

A large number of metals and non-metals may be evaporated by 
these means, provided they do not react with tungsten or molybdenum, 
and details of some are given in Table 8.1 (see also Olsen, Smith and 
Critenden*^'). 

Table 8.1. Evaporation Data for some Common Materials 







Approx. evaporation temperature 


Material 


Technique 


CC) 




p = 10-« torr 


p = 10-2 torr 


Aluminium 


Tungsten coil 


800 


1,100 


Carbon 


Arc 


2,000 


2,700 


Cobalt 


Tungsten coil or 








alumina crucible 


1,100 


1,650 


Copper 


Tungsten basket or 








molybdenum boat 


900 


1,300 


Gold 


Tungsten coil or 








molybdenum boat 


1,000 


1,500 


Iron 


Tungsten coil or 








alumina crucible 


1,100 


1,500 


Nickel 


Tungsten coil or 








alumina crucible 


1,100 


1,500 


Platinum 


Tungsten coil 


1,500 


2,100 


Silver 


Tungsten coil or 








molybdenum boat 


700 


1,050 


Zinc 


Tungsten basket or iron 








crucible 


200 


350 


Cryolite (NaaAlFe) 


Molybdenum boat 


500 


1,000 


MgF^ 


Molybdenum boat 


1,000 


1,500 


SeOa 


Molybdenum boat 


1,000 


1,500 


ZnS 


Molybdenum boat 


400 


500 



An alternative mode of evaporation useful for materials which do 
react with tungsten and molybdenum, is from a ceramic crucible, 
round which is wound a tungsten or tantalum heating element. 
The crucible, however, is a copious source of gas and needs lengthy 
degassing before evaporation can be commenced. 

174 



VACUUM COATING 

Another form of crucible is a blind hole drilled in a carbon rod 
which is heated by passing an electric current through it. 

The flash filament technique is a further method of evaporating 
materials which react with a heating element. The evaporant, in the 
form of fine powder, is allowed to fall towards the heater, which may 
be a molybdenimi strip heated electrically to about 2,000°K. The 
fine particles are heated to their evaporation point before actually 
coming in contact with the heater. 

A method of evaporation which avoids contamination originating 
from the crucible, is electron beam heating"'. In the simplest form 



Pendant drop 



Cathode—- 
loop 4 




Metal rod (anode) +20 kV 



Electron paths 



Focusing electrode 
-500 V 



bstrate 



Figure 8.4. Electron beam evaporator 



of this equipment, illustrated 'm Figure 8.4, the evaporant is in the form 
of a thin metal rod and is bombarded by high velocity electrons. The 
electron source is a loop of tungsten wire heated to about 2,000°K, and 
focusing is provided by a cylindrical electrode at a- potential of about 
— 500 V. The metal rod is maintained at a potential of up to 20 kV 
and electron currents up to 1 00 mA are used. The electron bombard- 
ment causes the tip of the rod to melt and form a pendant drop from 
which evaporation takes place. 

8.2.3 Sputtering 

The metal is either formed into a disc several inches in diameter, or 
it may be electrodeposited on a copper disc. The disc is suspended 

175 



SOME APPLICATIONS OF VACUUM TECHNIQUE 

directly above the substrate and within a metal ring of diameter 
slightly greater than that of the disc. A potential difference of 
several kilovolts is maintained between the disc and the ring, the disc 
being the negative electrode. An inert gas, usually argon at a pres- 
sure of about 10"^ torr, is leaked into the system through a needle 
valve and a glow discharge occurs between the disc and the ring. 
The positive ions of argon bombard the disc and sputter the cathode 
material which then collects on the substrate to form a fihn. The 
rate of sputtering increases as the gas pressure rises. The advantage 
of sputtering over evaporation is that it is a slow process and hence 
allows accurate control of film thickness. Its main disadvantage is 
that some gas becomes sorbed within the film. 

Thin films of metallic oxides may be produced by reactive sputter- 
ing, in which the metals are sputtered in an atmosphere consisting of 
about 5 per cent oxygen and 95 per cent inert gas. The oxide is 
formed either directly at the cathode before being sputtered or during 
the passage of the metal through the gas . These films of oxide adhere 
very strongly to glass surfaces. 

8.2.4 Thickness Measurement and Monitoring 

The thickness of deposited films may be estimated by a number of 
methods, including : 

{a) Determination of the weight and area of the deposit'^'. 

[b) Measurement of the electrical resistance of conducting films<^\ 

(c) Determination by radioactive tracers of the fraction of the 
original evaporant deposited*^°l 

{d) Interferometric methods'^^'. 

(«) Measurement of the optical transmittance'^^'. 
It is also possible to measure the film thickness during deposition. 
One method of doing this is by the measurement of the optical 
transmittance of the film as in {e) above, so that when the transmit- 
tance reaches a predetermined value the process is terminated. A 
second method of monitoring is by means of a quartz crystal exposed 
to the evaporant. The resonant vibration frequency of the crystal 
depends on the mass of deposit and when a preset frequency is reached 
the process may be stopped. 

8.2.5 Examples of Vacuum Coatings 

8.2.5.1 Aluminizing of Mirrors — Mirrors for precision optical instru- 
ments are invariably of the front reflecting type and are formed by the 
deposition of an aluminium film. 

The mirrors are mounted in a vacuum chamber and the aluminium 
evaporated from tungsten spirals at a pressure of about 10"* torr. 

176 



VACUUM COATING 

Plant suitable for this purpose ranges from small laboratory units to 
those with chamber sizes of 72 in. diameter in which a large number of 
mirrors can be aluminized simultaneously. The smallest of these 
units has a bell jar of about 4 1. volume, and is pumped by a 1 in. 
oil vapour pump of speed 10 1. sec"-'^ backed by a rotary pump of 
speed 30 1. min"^. On the 72 in. unit the working volume is about 
4,000 1., the total vapour pump speed 9,600 1. sec"^ and the rotary 
pump speed 7,500 1. min " ^. A 72 in. unit is shown in Figure 8.5, the 
chamber being pumped by six 9 in. vapour booster pumps in parallel, 
backed by a single rotary pump. 




Figure 8.5. A 72 in. coating unit {by courtesy o/ Edwards High Vacuum Ltd.) 



Large mirrors for astronomical telescopes are generally produced 
in specially designed chambers. The diameter of the chamber is just 
greater than that of the mirror and many evaporant sources are iwed 
to ensure a uniform deposit. A typical assembly for an 84 in. 
diameter mirror would be pumped by eight oil vapour pumps in 
parallel, each of speed 1,000 1. sec"^. 

After deposition, and on exposure to air, the aluminium develops a 
thin film of oxide which does not significantly affect its reflecting 
power but which acts as a protective layer against abrasion. Addi- 
tional protection may be provided by evaporating over the aluminium 
a thin transparent layer of magnesium fluoride or silicon oxide*'^^'. 



12- 



177 



SOME APPLICATIONS OF VACUUM TECHNIQUE 

8.2.5.2 Metallized Paper— Metal films deposited on thin paper or 
plastic are used (J) in the manufacture of electrical capacitors, and 
(ii) for heat insulation. 

The manufacturing process calls for the deposition of the metal on a 
roll of paper and hence for the continuous evaporation of the metal. 
Zinc is the metal most frequently used<"> as it evaporates at a low 
temperature. In order to obtain good adhesion of the zinc it is 
necessary to degas the paper of water vapour and also to coat the 
paper with a thin film of silver. The silver is evaporated from an 
electrically heated trough-shaped crucible over which the paper is 
drawn. A much thicker layer of zinc is then evaporated from a 
similar crucible. 

Aluminium has also been used for continuous deposition although 
the reaction of the aluminium with the crucible presents a serious 
problem'^^'. 

As large quantities of gas must be handled the pumping is 
carried out by vapour booster pumps backed by rotary pumps of 
suitable speeds. 



8.2.5.3 Decorative Coatings — MetaUic films may be deposited on plas- 
tics to give decorative and protective finishes. The surface of the 
plastic must be smooth and free from scratches; this is achieved by 
applying a coat of lacquer which also seals the plastic and reduces its 
outgassing. 

A unit for the vacuum metallizing of plastics is described by 
Barker<i^>. The articles to be metallized are loaded on circular 
spools which can be rotated during processing. The base lacquer is 
sprayed on and the spool and its contents are placed in an oven at 
about 55°C to stove the lacquer. The spool is then transferred to the 
vacuum plant where the metal is deposited in a chamber similar to that 
shown in Figure 8.5. Finally, a second layer of lacquer is sprayed 
on and stoved. This layer is essential to protect the metal layer and 
it may also be coloured for additional decoration. 

The pressure needed for evaporation is about 5 x 10""* torr and this 
is achieved by two vapour diffusion pumps in parallel (total speed 
10,000 1. sec"^) backed by a vapour booster pump and then a rotary 
oil pump. 



8.2.5.4 Corrosion Resistance Coatings— Ihic^i fihns of aluminium 
(0-001 in.), cadmium (0-0005 in.) or nickel (0-0005 in.) vacuum 
deposited on steel give good protection against corrosion by salt or 
fresh water'^''*. Thicknesses of this order require evaporation times 

178 



VACUUM COATING 

of about half an hour and are thus expensive. The type of chamber 
shown in Figure 8.5 may also be used to deposit these films. 

8.2.5.5 Optical Films — Thin films of non-metallic materials can be 
used for (i) the blooming of optical components to reduce light losses, 
and {ii) the production of interference and heat filters. 

The theories of these devices have been given by several authors 
{e.g. Ward*-'^'). It is necessary to deposit accurately successive 
layers of dielectrics each one-quarter of a wavelength thick, and this 
may be accomplished by one of the monitoring techniques described 
in Section 8.2.4. The materials used are ZnS, MgFa, cryolite and 
Ti02, all of which can be evaporated from molybdenum boats. 
These processes can be carried out in plant similar to that used for 
aluminizing mirrors. 

8.2.5.6 Microcircuits — The production of components for micro- 
circuitry can be achieved by the deposition on a glass substrate of 
suitable layers'-'-^'. Electrical resistors can be made by the deposition 
of narrow strips of a material such as Nichrome'^"^ ; a mask interposed 
between the evaporation source and the substrate determines the 
exact shape of the strip. End contacts to the resistance strip are 
made by evaporating thick layers of copper or aluminium, using 
further masks to cover the rest of the substrate. It is thus necessary 
to use at least two evaporant sources which are brought into the 
evaporation position by a rotating jig which also operates the ap- 
propriate mask. 

Capacitors are made by evaporating alternate layers of metal and 
dielectric'^^\ As the dielectric films can be made thin, high values 
of the capacitance per unit area can be obtained although the 
working voltage is low ( ~ 60 V) . An additional cause of electrical 
breakdown is dust and surface irregularities on the metal which lead 
to thin patches of dielectric material. The metal layers usually con- 
sist of aluminium about 500 A thick and the dielectric layers are of 
silicon monoxide about 10,000 A thick. A multiple-source rotating 
jig is used to evaporate these materials, the dielectric in powder form 
being contained in molybdenum boats, and the aluminium in wire 
form on tungsten spirals. A coating unit such as that described for 
aluminizing mirrors is also suitable for this purpose. 

Complete microcircuits can be deposited on small glass plates 
by using suitable masks and a multiple evaporation head*^^'. The 
resistors are deposited first, followed by the lower plates of the 
capacitors together with some of the connecting strips. To avoid 
undue multiplication of sources all these layers are generally of 

179 



SOME APPLICATIONS OF VACUUM TECHNIQUE 

nickel-chromium alloy which can be evaporated from a tungsten 
wire. The dielectrics are deposited next, and then the top plates of 
the capacitors and the remainder of the connecting strips, again all in 
nickel-chromium alloy. Finally, thick copper deposits are put down 
as soldering points. 

Microcircuits are also produced in a number of interconnected 
chambers, one deposition process taking place in each chamber, thus 
avoiding the complications of multiple-source rotating jigs. 

After removal from the vacuum plant, transistors are soldered in 
position to complete the circuit. 

8.2.5.7 Computer Elements — Another use of evaporated films in 
electronics is as storage elements in computers. Films of magnetic 
materials are used and in order to obtain consistent properties it is 
necessary to deposit the films at a pressure of about 10"" torr. This 
pressure can be obtained by an oil vapour pump and a liquid nitrogen 
trap, or by sputter-ion pumps. 

8.2.5.8 Thin Film Research — The study of the optical, electrical and 
magnetic properties of thin films is a major field of solid state physics, 
as the quasi two-dimensional atomic array of thin films presents a 
great simplification of the general three-dimensional properties of 
normal materials. 

This field of research has brought about (t) the use of ultra-high 
vacua to reduce surface contamination, and {ii) the deposition of 
single crystal (epitaxial) films. The first of these developments has 
led to the production of ultra-high vacuum coating units, using either 
vapour diffusion pumps and cold traps or getter-ion pumps. Epi- 
taxial films are produced by allowing the evaporated material to 
condense on a heated crystalline substrate*^'^'. 

8.2.5.9 Electron Microscopy Specimens — Two types of thin specimens 
suitable for use in electron microscopy may be produced by vacuirai 
evaporation: 

(a) Thin films are used for studies by direct electron transmission 
and electron diffraction, and hence very thin specimens ( ~ 200 A) are 
required. The technique adopted is to deposit the fihn by evapora- 
tion on to a substrate such as rocksalt, the latter being then dissolved 
away in water. The film floats to the surface of the water and may be 
collected on the standard copper grids used in electron microscopy. 

{b) Shadowed specimens are used for viewing surface structure and 
irregularities. The first step is to make a replica of the surface with a 
material such as cellulose acetate. Then carbon is evaporated to fall 

180 



TUNGSTEN FILAMENT LAMPS 



obliquely on the replica, so that surface irregularities throw a carbon- 
free 'shadow' which can be observed when the replica is viewed in the 
electron microscope. 

The carbon is deposited by striking an arc between two thin carbon 
rods, the heat of the arc causing evaporation. 



8.3 Tungsten Filament Lamps (GLS lamps) 

These consist of a tungsten filament enclosed in a glass envelope 
filled with inert gas. It is essential that the oxygen, water vapour and 
hydrocarbon content of the finished lamps be made as small as pos- 
sible, to avoid premature failure due to reactions with the hot tung- 
sten. A major source of water vapour is that sorped in the glass of 
the envelope, whereas that of oxygen is the normal volume gas. To 
remove these gases by conventional means would require long pump- 
ing times at fairly high speeds. However, economic considerations 
dictate that the pumping time should be as short as possible, and in 
addition the physical construction of the lamp requires the exhaust 
tube to be of small bore thus precluding high effective speeds. 

These requirements have been reconciled in a technique whereby 
the envelopes, having been heated, are pumped for a short time and 
then filled with nitrogen. This process is repeated several times, with 
finally a longer evacuation followed by filHng with inert gas. Thus 
most of the water vapour is removed by the enhanced outgassing at 
the elevated temperature and the oxygen is replaced by nitrogen. 
The circulation of the nitrogen is also thought to have a scrubbing 
action on the walls of the envelope thus assisting in water vapour 
removal. The presence of residual nitrogen is of little consequence 
since nitrogen is a component of the gas with which the lamp is 
eventually filled. A final clean-up of oxygen is carried out by firing a 
getter within the completed and sealed lamp. 

The manufacture of such lamps is carried out semi-automatically by 
a unit consisting of three interconnected machines. The first of these 
is used to form the filament from suitably prepared tungsten wire, 
mount it in a glass bead and finally coat it with a thin solution con- 
taining red phosphorus. On the second machine the bead and a 
narrow evacuation tube are sealed into a previously formed envelope. 
During the process of transfer from the second machine to the third 
(exhaust) machine the envelope passes through a heated tunnel so 
that when it reaches the exhaust machine it is at a temperature of 
about 400°C. The third (exhaust) machine carries out the evacua- 
tion, flushing, filling and final sealing operations and, from the point 
of view of vacuum technique, is the most interesting. 

181 



SOME APPLICATIONS OF VACUUM TECHNIQUE 

The exhaust machine, consists essentially of some 30-36 heads 
arranged round the circumference of a horizontal circle of about 3 ft. 
radius. Each head terminates in a short length of rubber tube, into 
which the lamp evacuation tube is a tight fit, and is connected via a 
cam operated two-way valve to the pumping and gas filling lines. 
The pumping line from each head is of ^ in. bore copper tube and is 
connected via a manifold to a central rotary seal which in turn is 




Figure 8.6. Exhaust machine for manufacture of tungsten filament lamps {by courtesy 
o/ British Lighting Industries Ltd.) 



connected to a single rotary pump having a speed of about 3,000 1. 
min"-'^. The rotary seal is made between two metal plates which are 
accurately lapped together and fringe tested to ensure perfect flatness. 
Lubrication and vacuum sealing of the plates is carried out by castor 
oil. 

A typical exhaust machine is shown in Figure 8. 6. The cams for the 
valve mechanism may be seen in the centre foreground, whilst on the 
upper left may be seen the heating tunnel leading from the second 
machine. 

By virtue of the rotary seal the whole assembly of heads can rotate 

182 



DISCHARGE LAMPS 

in Steps and, in general, a different operation is carried out at each 
step or index position. 

At the first index position the hot lamp assembly is transferred from 
the second machine to a vacant head on the third where it is evacuated 
from atmospheric pressure to about 1 torr. It is then passed to the 
next position where it is isolated so that, if there is a large leak in the 
envelope, the pressure in the bulb will rise. On the following position 
the lamp is connected to a mercury manometer and, if the pressure 
has risen by more than a predetermined amount (depending on the 
size of the envelope) the mercury completes an electrical circuit which 
operates a mechanism for rejecting the lamp. 

The next twelve or so positions are used alternatively to evacuate 
and fill with dry nitrogen. The time spent on each position is only a 
few seconds during which the pressure is reduced to about 1 " ^ torr or 
filled to about 100 torr. As the diameter of the lamp evacuation tube 
is about 0-3 cm the gaseous flow over this pressure range is in the 
viscous and transitional regions (seeeqn2.13). The length of this tube 
is about 10 cm and thus the conductance falls from about 15 1. sec "^ at 
100 torr to about 0-02 1. sec"^ at 0-1 torr. Towards the end of the 
pumping cycle the effective speed in the lamp is governed entirely by 
the low tube conductance, and consequently little advantage would 
be gained by using a larger pump or by pumping for a longer time. 

After the alternate pumping and flushing the lamp is evacuated at 
several consecutive stations before it is filled with an argon-nitrogen 
mixture to a pressure of about 700 torr and sealed off. During the 
flushing and evacuation stages the lamp has cooled to about 80°C. 
The final filling temperature and pressure are such that under normal 
operating conditions the inert gas pressure is rather greater than 
atmospheric. 

The finished lamp is allowed to stand for a week, after which it is 
tested by applying to it the normal operating voltage. This has the 
effect of firing the phosphorus getter resulting in a final clean-up of 
oxygen. Should the lamp be slightly leaky, the amount of oxide 
produced is suflScient to form a deposit on the lamp walls which can be 
easily detected. 



8.4 Discharge Lamps 

In these lamps the light is emitted by an electrical discharge through a 
suitable gas contained within the lamp envelope. Typical examples 
are the fluorescent tube with its low pressure mercury vapour filUng, 
the sodium lamp and the high pressure mercury lamp used for street 
lighting. 

183 



SOME APPLICATIONS OF VACUUM TECHNIQUE 

The vacuum requirements of these lamps are similar to those of 
GLS lamps but the nitrogen flush technique does not seem to have 
been perfected for their production. Instead, the lamps are evacu- 
ated by oil vapour pumps to air pressures between 10 ~* and 10"'' torr 
after having the active element {i.e. mercury and sodium) added. 
Fluorescent tubes are sealed and evacuated on rotary indexing 
machines similar to but larger than those used for GLS lamps. 

The sodium lamp arc tubes are evacuated in groups of six to about 
10~* torr and sealed off". The arc tube is enclosed within an outer 
envelope which is internally coated with an infra-red reflecting 
film and this outer jacket is also evacuated to about 10 ~* torr. 
This heat conservation permits high light efficiencies to be achieved. 



8.5 Manufacture of Electron Tubes 

8.5.1 Radio Receiver Valves 

These devices consist of an electrode assembly mounted inside a 
glass envelope in which a pressiu'e of about 10"^ torr is required. 
The electrode assembly is usually made up of a central cathode 
surrounded by a number of wire grids and finally a sheet metal anode. 
The cathode consists essentially of a nickel former coated with barium 
and strontium oxides, which are formed in the manufacturing process 
by reduction of the carbonates, this process being known as cathode 
activation. 

The vacuum requirements are the production of a pressure of 
about 1 ~ ^ torr and the maintenance of this throughout the life of 
the valve in order to avoid contamination of the electrode surfaces, 
particularly that of the cathode. The initial pumping to 10 ~® torr is 
done by conventional means and the final reduction to 10"^ torr and 
maintenance at this pressure is by the use of a getter. 

The chief sources of gas during the pumping process are (?) water 
vapour from the glass envelope, (m) hydrogen from the nickel elec- 
trodes, {Hi) carbon dioxide from the cathode activation process, and 
{iv) argon during the firing of the getter. 

Figure 8.7 shows a typical machine which is used in making radio 
valves. The electrode assembly, having been sealed into a glass base 
on a separate machine (not shown) is then hand loaded on to a rotary 
table where the base is sealed into a glass envelope to which is already 
attached a narrow pumping tube. This operation may be seen on the 
right of Figure 8.7. The completed envelopes are transferred by a 
mechanical arm (centre) to the pumping table on the left of the 
picture. 

As in the manufacture of lamps the pumping and sealing are carried 

184 



manufacture of electron tubes 

out on a rotary machine but there are several important differences 
between the two techniques. In valve manufacture, the finished 
product must be under high vacuum so the ffushing process is not 
used. Instead the valve is pumped continuously on all the index 
positions. In older equipment each index position was connected 
directiy to a vapour pump and all the vapour pumps were backed by a 
single rotary pump via a rotary seal in the centre of the table. Modem 
equipment, however, generally uses a complete pumping unit of a 




Figure 8.7. Radio valve assembly and pumping {by courtesy o/MuUard) 



vapour pump (~ 701. sec"-') backed by a rotary pump (~501. 
min"^) to each index position, this arrangement simplifying main- 
tenance. The effective speed inside the valve envelopes is much 
less than that of the pump because of the high impedance of the 
stem. 

A rotary table of this kind has 25-30 heads each terminating in a 
collet type fitting to take the valve stem. The valve spends up to 
3 min at each index position, this time being determined by the longest 
process which is usually cathode activation. On the first ten or so 
positions the valve passes through an oven at a temperature of either 
400°C for hard glass or 250°C for soft glass envelopes; the purpose of 

185 



SOME APPLICATIONS OF VACUUM TECHNIQUE 

this heating is to degas the glass envelope. On emerging from the 
oven the valve is allowed to cool on two more positions. After this 
the cathode is activated by heating the tungsten filament electrically 
so that the cathode temperature rises to 1,200°C. During this process 
a large volume of carbon dioxide is evolved and sufficient time must 
be allowed for this to be pumped away. The next few positions are 
occupied in degassing the electrode structure which is carried out by 
eddy current heating. A radiofrequency coil drops over the valve 
and the electrodes are brought to red heat when hydrogen is liberated 
and pumped off. The valve is then operated at several times its 
normal operating conditions to bring about final degassing. At this 
point the pressure in the valve is between 10"^ and 10"^ torr. In 
order to bring about a final pressure reduction a barium getter is 
fired inside the envelope. As the barium may have been prepared in 
an atmosphere of argon, this process may liberate a considerable 
quantity of argon and when this has been pumped away the valve is 
sealed off. The action of the getter reduces the pressure to about 
10"^ torr and maintains this pressure throughout the life of the 
valve. 

8.5.2 Cathode Ray Tubes 

The vacuum requirements of these devices are basically the same as 
for radio valves and the techniques used in production are similar in 
principle. The glass envelope is prepared separately and the fluores- 
cent screen deposited chemically before the electron gun assembly is 
sealed in. The cathode ray tube is then attached to a pumping unit 
mounted on a trolley, the pumps having the same speeds as those 
used in valve manufacture (see Figure 8.7). The trolleys are carried 
round to the various processes either manually or by a moving belt, the 
sequence of operations being the same as in the case of valves. 

8.5.3 X-ray Tubes 

These require similar vacuum conditions to the cathode ray tubes 
and are produced on a similar trolley system. The x-ray tubes are 
ptimped and a high voltage applied across the electrodes until they 
spark over, the resultant ionization of the residual gas causing 
degassing of the electrodes by ionic bombardment. 

8.5.4 Voltage Stabilizers 

These consist of two electrodes in an inert gas, a gas discharge 
being produced when the voltage across the electrodes rises above a 
certain value. The envelope is pumped to 10"^ torr as described for 
radio valves and then filled with the gas at a pressure of 25 torr. 

186 



MANUFACTURE OF ELECTRON TUBES 

In order to produce a constant striking voltage it is essential that the 
electrodes should have clean surfaces. The electrodes are generally 
made of molybdenum and after pumping and sealing off the valve is 
overrun. The resulting discharge sputters molybdenum from the 
cathode and leaves a clean surface. The sputtered molybdenum also 
acts as a getter in cleaning up any active components in the inert gas. 

8.5.5 Travelling Wave Tubes 

These devices are used for producing radio waves of frequencies 
greater than 10^° csec ~ ^ and consist of an electron gun firing a beam of 
electrons along the axis of a tungsten spiral on to a copper anode. A 
radio wave passes along the spiral and the electrons are bunched, thus 
providing high frequency pulses at the anode. At the high frequen- 
cies used the radius of the helix is very small and the whole device is 
very narrow thus presenting a large pumping impedance. It is 
necessary to operate the device at a pressure of about 10 ~^ torr to 
prevent the scattering of electrons and hence ultra-high vacuum 
technique is needed'^*'. 

Each device receives its final pumping by an evapor-ion pump 
with a speed of about 0-4 1. sec" ^, which can also be used as an Alpert 
gauge to measure the pressure. An evapor-ion pump is chosen in 
preference to a Penning pump because the magnetic field of the latter 
would deflect the electron beam when the device is being operated 
during pump-down. 

The devices with their evapor-ion pumps attached are prepumped 
in batches of up to eight on a common vapour pump unit. The 
devices are sealed into a branched stem each branch being about 
1 \ in. in diameter and 2 ft. long. The stem is connected to an oil 
vapour pump via a liquid nitrogen trap and the whole system is 
pumped by the vapour pump to a pressure of 10"^ torr (as recorded 
by the Alpert gauge) . At this point the devices are baked by having 
their temperature raised slowly to 400°G and held there for 48 h, this 
process removing water vapour from the glass and the cathode binder. 
After cooling, the cathode is activated as before by heating it to 
1,200°C and then the electrodes are degassed by electron bombard- 
ment from the cathode. At the same time, the electrodes in the ion 
pump are also degassed and also the glass neck of the tube at which 
the sealing off process will take place. The pressure should then be 
about 10"'^ torr. 

The device is then sealed off from the vapour pump and the evapor- 
ion pump takes over the pumping action and reduces the pressure 
to 10 ~® torr. The devices are screened, i.e. run under full operating 
conditions for up to 5 days, and finally sealed off from the ion pump. 

187 



SOME APPLICATIONS OF VACUUM TECHNIQUE 

After storage the tubes are leak tested by using their electrode 
structure as an ion gauge to detect any significant pressure rise. 



8.6 Vacuum Metallurgy 

8.6.1 Introduction 

The use of vacuum techniques in metallurgical processes is ex- 
tremely diverse to the extent that most of these processes have been 
carried out in a vacuum environment. Vacuum processing is more 
expensive than conventional practice both in the initial cost of plant 
and in its day-to-day rurming, but generally the improvement in the 
final product is well worth the extra cost. 

In order to determine the usefulness of a vacuimi process, and the 
conditions of temperature and pressure that must be achieved for the 
process to be successful, the kinetics and thermodynamics of the 
reactions involved must be studied. These considerations are outside 
the scope of this book but have been adequately discussed by several 
authors (for example Belk'^^' and King<^^^). 

From the vacuum point of view the three main problems are [i) the 
very considerable evolution of gas during most processes, thus neces- 
sitating very high pumping speeds, (m) the contamination of vessels, 
pumps, pump fluids, etc., by the gases and vapours evolved, and {iii) 
the production and maintenance of high temperatures in a vacuum 
environment. 

The vacuum equipment used ranges from comparatively small 
systems which are used in metallurgical research, small-scale produc- 
tion of new and expensive materials and pilot plant studies, to 
extremely large plant used for the production of tonnage quantities. 
This wide range of plant makes it very difficult to quote general 
examples, but wherever possible, typical examples will be given in the 
relevant sections. 

8.6.2 Vacuum Degassing 

In this process metal is melted in conventional (atmospheric) 
furnaces and then passed into a vacuum environment where it is 
degassed. The degassed metal may then be cast in air or (preferably) 
in vacuum. The advantages of such a process compared with the 
normal one are that (i) the product is considerably less brittle (due to the 
removal of hydrogen), («) ductility is improved, {in) a much smaller 
quantity of oxide inclusions (which are individually smaller and more 
uniformly distributed) is achieved, and (w) electrical and thermal 
conductivity are improved. 

188 



VACUUM METALLURGY 

Since it is not necessary to provide melting facilities within the 
vacuum it is possible to handle very large quantities of metal and the 
process has been applied extensively in the production of steel, copper 
and aluminium. Three basic methods of degassing are employed. 

8.6.2. 1 Static Degassing — A ladle of molten metal is placed in a chamber 
which is then sealed and evacuated. A typical plant will degas 25- 
35 ton at a mean pressure of about 30 torr for 10 min. The dis- 
advantages of this relatively simple method are that (t) since the metal 
is not heated whilst it is under vacuum the degassing time must be short 
to avoid solidification before pouring, (m) the metal is usually cast 
in air which will re-introduce some gas, and (m) degassing takes place 
only from near the surface since ferro-static pressure prevents bubble 
formation at depth. The last of these disadvantages can be overcome 
by producing a stirring action with a stream of inert gas forced up 
through the liquid, but this, of course, increases the total gas load very 
considerably. 

8.6.2.2 Progressive Degassing — ^A portion of the molten metal is auto- 
matically transferred from the furnace to a vacuum chamber, de- 



To vacuum 
pumps 



To vacuum 
pumps 




Figure 8.8. Vacuum lift de- 
gassing : (a) return to crucible, 
(b) degassing 



•) > > ) I f 1 1 ^ J J } } } t;^ Ladle or v/ >>>/.'>>> r >>>> ^ 
melting crucible 

(a) (63 



gassed, and then returned to the furnace. This process is repeated 
until the desired degree of degassing has been achieved. 

189 



SOME APPLICATIONS OF VACUUM TECHNIQUE 

One method of doing this is the vacuum lift method, which is 
illustrated in Figure 8.8. The vacuum vessel is connected to the melt- 
ing crucible by a wide bore pipe such that atmospheric pressure 
causes the metal to rise into the vacuum vessel and a large surface is 
exposed from which degassing takes place. After a suitable time the 
vacuum vessel is raised and most of the metal falls back into the cru- 
cible giving a stirring action. Typically, the lift and fall process is 
repeated about thirty times. 

To vacuum 
pumps 



Splash 
baffle 



Flushing 
gas 




Figure 8.9. Circulation degassing 
{after Belk«6>) 



~55 in. 
for steel 



An alternative method is circulation degassing in which a vacuum 
vessel is connected to the crucible by two pipes as illustrated in 
Figure 8.9. Metal rises through both pipes to expose a large surface 
within the vacuum. A stream of argon is forced into one pipe which 
blows the molten metal into the vacuum vessel in the form of small 
droplets and causes a circulation of metal from the crucible through 
the vacuum vessel. The process may be continued for any desired 
time. 

Both the lift and the circulation methods are used to degas charges 
of up to 80 ton in 15-20 min, with pressures ranging from 100 torr 

190 



VACUUM METALLURGY 



at the beginning of the process to about 5 torr at its completion, 
both methods the metal must be cast in air. 



In 



8.6.2.3 Stream Degassing — ^The crucible of molten metal is placed 
above a vacuum chamber containing either a pouring ladle or one or 
more ingot moulds. The crucible is connected to the vacuum 
chamber via a small plugged hole. When the plug is removed the 
metal streams into the vacuum chamber and into the ladle or mould 
in such a way that it breaks into small droplets which ensure a reason- 
ably efficient release of gas. Pouring rates are usually of the order of 
7-10 ton min"^ producing ingots of up to 200 ton'^'". A disadvan- 
tage of this method is that the degassing time is short and not readily 
controllable, but it is offset to some extent by the fact that the metal 



To vacuum 



pumps 




Ingot 
moulds 



Figure 8.10. Vacuum degassing and casting [after Belk'^^') 



can be cast in vacuirai. An improvement in degassing can be made 
by carrying out one of the progressive methods before stream degas- 
sing. A typical plant for multi-ingot casting is illustrated in Figure 
8.10. 
Typical purification given by the above methods of degassing is : 

Hydrogen content reduced from an initial 3-7 ppm to 0-5-2 
ppm. 

Nitrogen content reduced to about one-third of its initial value. 

Inclusion content, in terms of total oxide content, reduced to 

0-04 per cent for air cast material and to 0-01 per cent for vacuum 

cast. 

Since the pressures required for degassing are in the 5-100 torr range, 

Roots pumps backed by suitable oil sealed rotary pumps are adequate. 

The average gas load is typically 100 torr 1. sec~^ per ton of material, 

191 



SOME APPLICATIONS OF VACUUM TECHNIQUE 

but at the commencement of degassing may be up to ten times as 
great. 

8.6.3 Vacuum Induction Melting 

This technique has been developed particularly to enable metal to 
be melted entirely under vacuum and to allow very precise control of 
alloy composition, especially when using readily oxidizable alloying in- 
gredients. One disadvantage of the method is that because the whole 
furnace is under vacuum the quantities which can be handled are 
small compared with those in vacuum degassing. A second dis- 
advantage is that normal refractory crucibles (usually of zirconia, 
magnesia-alumina or graphite) must be used, and reactions between 
the crucible and the molten metal proceed much more rapidly under 
vacuum, resulting in contamination of the product and rapid destruc- 
tion of the crucible. 

The crucible and mould into which the metal is eventually poured 
(either by bottom pouring or crucible tilting) are contained in a 
vacuum vessel whose walls are water-cooled. The crucible is sur- 
rounded by a coil carrying an alternating current and the eddy 
currents set up in the metal produce sufficient Joule heating to melt the 
metal. 

Typical furnaces range from a few pounds capacity using some 50 
kW of power to 400 V and 5 kcsec " S to 5,000 lb. capacity using 1 , 1 00 
kW at 1 kcsec ~^*^^'. The working pressure is about 10"^ torr, the 
main pumping being performed by vapour booster pumps. The 
smaller furnaces would use a 6 in. pump {S ^ 600 1. sec"^) whereas 
the largest furnaces use twelve 16 in. pumps in parallel (total speed 
~ 60,0001. sec"-'^). Backing pump speeds are chosen rather higher 
than are strictly necessary for the booster pumps so that they can deal 
rapidly with sudden bursts of outgassing which sometimes occur, 
especially when alloying materials are added. 

An example of a system of intermediate size is the plant described by 
Wright and Harper'^^' which was designed primarily for the melting 
of uranium and its alloys. The melting capacity is 200 kg of uranium 
(bulk equivalent to 200 lb. of steel) and main pumping is provided by a 
30 in. vapour booster pump with a speed of 4,500 1. sec"-' at 10"^ torr. 
In addition, two 16 in. vapour pumps in parallel are provided to give 
working pressures down to 10"* torr for extreme purification and also 
to assist the booster pump if necessary. Roughing and backing is 
provided by an 8,700 1. min ~ ^ Kinney pump. A 9 in. vapour booster 
pump is installed in the bacldng line of the vapour pumps so that the 
vapour pumps can operate in parallel with the main booster pump. 
Introduction heating power is 100 kW at 2,850 csec"^. Figure 8.11 

192 



VACUIBI METALUmOY 

gives a general view of the furnace, pumping manifold and the 
main pumps. The main booster pump can be seen in the fore- 
ground. 

Apart from handling the gas load the vacuum problems associated 
with this type of furnace include the provision of rotating or sliding 




Figure 8.11. Overhead view of the pumping system of the 200 kg bottom pouring 
uranium furnace. The lid is removed to show the interior and crucible assembly {by 
courtesy of Edwards High Vacuum Ltd. and The Australian Atomic Energy 

Commission) 



seals to operate the pouring mechanism and the provision of vacuum- 
tight high current electrical lead-throughs. In addition, it is common 
practice to provide a vacuum lock through which raw material, 
alloying ingredients, etc., can be added to the melt without breaking 
the vacuum in the main chamber. Such a lock usually consists of a 
subsidiary chamber mounted above the main one with a hinged lid to 



13 — I.H.V.T. 



193 



SOME APPLICATIONS OF VACUUM TECHNIQUE 

the atmosphere and a hinged plate to the main chamber. Figure 8.12 
illustrates a bottom pouring furnace with a vacuum lock. 



VACUUM METALLURGY 



To lock 
pumps 



To main 



pumps 




Figure 8. 1 2. Induction heated 

furnace with vacuum lock {after 

Belk"") 



Graphite 
crucible 

High frequency 
coils 



Removable plug 



Water cooled 
mould chamber 



8.6.4 Vacuum Arc Melting 

This process has been developed for melting those metals which are 
so reactive that they can neither be melted in air nor contained in a 
refractory crucible. Typical of such metals are molybdenum, 
tungsten, niobium, zirconium, and titanium. The two forms of arc 
melting used are («') the consumable electrode, in which the metal to 
be melted acts as one electrode, and (if) the permanent electrode for 
metals and alloys which cannot be readily formed into the appropriate 
electrode shape. 

8.6.4.1 Consumable Electrode Furnaces — ^The material to be melted is 
formed into a cylindrical shape and initially placed in a water-cooled 
mould of a suitable metal (copper is frequently used), the mould being 
positioned beneath an open bottomed cylindrical vacuum chamber. 
The top of the electrode is then mechanically and electrically at- 
tached to a metal rod which passes vertically up through the chamber 

194 




Figure 8.13. The electrode and ingot mould of a 

consumable electrode arc furnace being raised into the 

vacuum chamber [by courtesy of Imperial Metal 

Industries Ltd.) 



195 



SOME APPLICATIONS OF VACUUM TECHNIQUE 

and through a sliding seal of the Wilson type to external lifting gear. 
The mould vessel is raised and sealed to the chamber and the whole 
system evacuated. An arc is then struck between the electrode and a 
few chips of electrode material lying at the bottom of the mould. As 
the electrode melts its position is adjusted by the Ufting gear so that 
conditions for optimum arc between the electrode and the surface 
of the melt are maintained. 



Rod control 
gear 



Shaft seal 



Control 
rod 



Consumable 
electrode 




Figure 8.14. Consumable electrode arc 
furnace 



To vacuum 
pumps 



Cooling 

water 

out 



Cooling water In 



Figure 8.13 shows the mould vessel of a typical consumable elec- 
trode arc furnace being raised towards the vacuum vessel. The 
electrode, of sintered titanium, may also be seen. Details of the 
construction of such a furnace are illustrated in Figure 8.14. Working 
pressures are of the order of 10"^ torr or less, the pumping being 
provided by vapour booster pumps backed by Roots ptimps which, in 
turn, are backed by rotary pumps. Furnaces range from those 
pumped by one 6 in. booster which produce 100 lb. ingots, to those 
giving 5,000 lb. ingots of 20 in. diameter, which require a pumping 

196 



VACUUM METALLURGY 



speed of 30,000 1. seC"^ and which consume some 1,000 kW at 
30-40 V. 



8.6.4.2 Permanent Electrode Furnaces — ^The construction of these 
furnaces is similar to that of the constimable electrode type. The 
electrode is usually of tungsten or graphite and the raw material (in 
the form of chips or powder) is fed to the arc from a hopper. The 
main disadvantage compared with the consumable electrode process, 
is that the gradual erosion of the electrode causes contamination of 
the melt. In order to minimize this erosion it is necessary to work at 
higher pressures ( ~ 1 torr) giving a smaller degree of degassing. This 
process is not so widely used, nor for such large capacities, as the 
constmiable electrode process. 

The vacuum problems associated with both processes are mainly 
those of providing a sufficiently high pumping speed at the arc to 
maintain the required pressure. This is particularly difficult in the 
early stages of melting since, in order to keep the overall height of 
the plant within reasonable bounds, the electrode diameter is made 
only slighdy less than that of the mould. The long annular gap 
between the electrode and the mould presents a large impedance to 
flow at the beginning of the process, but this impedance decreases as 
the molten surface and the arc move up the mould. The problem 
can be overcome by providing the mould with a retractable base, 
which is initially set near the top of the mould. As melting proceeds 
the base is lowered, thus ensuring that the molten surface remains 
near the top of the mould, with the result that the length of the 
annular gap is always small. 

Important vacuum seals are those associated with the electrode 
control rod and any vacuum locks which may be incorporated for the 
addition of raw materials and alloying ingredients. 

8.6.5 Electron Beam Melting 

The growing demand for high purity refractory metals {e.g. 
tantalum, tungsten, niobium) has led to the use of a beam of electrons 
as an almost contamination free heat source of extremely high tempera- 
ture capabilities. In principle, electrons are generated thermioni- 
cally and accelerated and focused by a set of electrodes; they then fall 
on the metal to be melted, where liie kinetic energy of the electrons 
is converted into heat. The metal is fed into the electron beam 
in the form of either a consumable rod or as chips or powder. 
As the metal melts it falls into a water-cooled metal ingot mould, 
the top siirface of the ingot being maintained in the molten state by 
further electron bombardment. The whole process takes place in a 

197 



SOME APPLICATIONS OF VACUUM TECHNIQUE 

low pressure environment and considerable degassing of the metal 
occurs. 

The simplest electron source consists of a ring cathode of pure 
tantalum or tungsten. Surrounding the cathode is an 'open-box' 
section accelerating-focusing electrode maintained at a high negative 
voltage with respect to the target. The shape of the box section 
determines the form of the resulting electron beam. Two possibili- 
ties are illustrated in Figure 8.15, which shows the general layout of an 
electron beam furnace. 



_ To electrode seal 
and control gear 



Consumable 
electrode 

Electron 

focusing 

boxes' 



-20kV 




Figure 8.15. General layout of an electron 

beam furnace using a consumable electrode and 

ring cathodes 



To vacuum 
pumps 



Water cooled 
mould chamber 



The main disadvantages of the ring cathode source are : 

(a) The cathode is easily contaminated, and its electron emission 
thus reduced, by its proximity to the target. 

{b) The whole of the chamber must be kept at pressures no higher 
than 10 ~* torr to avoid serious loss of electron energy and direction 
by collision with gas molecules. 

{c) The cathode power dissipation is limited to about 10 kW. 

The main advantages are that the electrode structure is simple, 
and can be easily and cheaply replaced. 

A more refined electron source consists of an electron gun in which 
the cathode may be an indirectly heated block of tantalum (shaped to 

198 



VACUUM METALLURGY 

give elementary focusing) permitting the use of powers up to 250 kW. 
A separate system of accelerating and focusing electrodes produces a 
constant velocity beam which means that the whole gun assembly can 
be remote from the target, thus avoiding cathode contamination. 
Further, the final accelerating electrode can be in the form of a 
small aperture which provides sufficient pimiping impedance for the 



-20 kV 

'. V \ , , .. 

■ l\ 

gun pumps- 
Focusing coils 

Vacuum lock 



Accelerating 
electrode and 
vacuum impedance 



To main 
chamber pumps 



vacuum pumps 



Raw material 




^Water cooled 
mould chamber 



Retractable base 



Figure 8.16. General layout of an electron beam furnace using an electron gun 



pressure in the electron gun to be significantly lower than that in the 
melting chamber. Such a system is illustrated in Figure 8.16. 

The practical upper limit to the accelerating voltage is about 20 kV, 
since at higher voltages the hazard due to x-rays produced at the 
target becomes serious. 

Typical plants range from a 60 kW furnace pumped by a 20 in. 
vapour pump (speed ~ 6,5001. sec" ^) and backed by a vapour 
booster pump and a mechanical rotary pump, to a 260 kW furnace 
with an effective pumping speed of 36,000 1. sec"^ in the working 

199 



SOME APPLICATIONS OF VACUUM TECHNIQUE 

pressure range of 10"*-10~^ torn This latter plant is capable of 
producing ingots up to 1 , 1 00 lb. in weight with diameters from 2 in. to 
7 in. and up to 72 in. in length'^"'. Smith, Hunt and Hanks*^^' 
describe a 1,000 kW furnace having a pumping speed of some 175,000 
1. sec"^ at 10 ~* torr which produces ingots up to 20 in. in diameter 
and 80 in. in length. 

8.6.6 Purification by Distillation 

At low pressures the volatilization of many metals is possible 
at temperatures of the order of 1,000°C and hence purification 
by fractional distillation can be achieved. Examples of its use 
are the reclamation of scrap brass by the distillation of the zinc 
content, the reclamation of aluminium alloy scrap, and the removal of 
manganese, magnesium, copper and sulphur from iron. Distillation 
has also been used in the production of magnesium from ore. The 
theory of the process has been reviewed by St Clair'^^'. A distilla- 
tion furnace consists essentially of a crucible above which is a condensa- 
tion tower with long sloping sides. The crucible and tower are heated 
either by radiant or induction heating. It is usually arranged that a 
temperature gradient is set up in the tower with the lowest temperature 
at the top. In this way separation between materials of similar 
volatility is obtained, the most volatile freezing only at the top of the 
tower. Pumping is carried out by vapour booster pumps and 
pressures of the order of 10~^ torr are used. 

8.6.7 Solid Phase Applications 

Although the main applications of vacuum in metallurgical 
techniques lie in the melting and purification of metals, its use in the 
treatment of metal in its solid form is significant. Of these processes 
those involving heat treatment are the most important. Obviously, 
heat treatment of the reactive metals must be carried out at low 
pressures, but many other metals benefit from vacuum treatment. 
For example, in the annealing of copper and high carbon stainless 
steels, the use of vacuum is advantageous in that the carbon pick-up, 
decarburization, oxidation and nitridation often associated with 
normal furnace atmospheres are avoided. 

Vacuum heat treatment furnaces can be externally heated by the 
usual gas or oil fired system but the poor thermal conductivity of the 
vacuum leads to inefficiency. For this reason internal heating 
methods {e.g. radiant or induction) are preferred. Pressures in the 
region 10"^-10~* torr are used. 

In a system for the continuous annealing of metal strip'^s)^ the strip 
enters the vacuum system via a mercury seal, passes through the 

200 



VACUUM METALLURGY 

annealing fiirnace, then through a cooling region, and finally re- 
enters the atmosphere through a second mercury seal. Pumps at 
each end of the plant maintain a mean pressure of the order of 10~^ 
torr. 

An alternative method which dispenses with the need for the 
mercury inlet and outlet seals is described by Krall, Scheibe and 
Ogiermann^^*' and is illustrated in Figure 8.17. The strip enters the 
anneaUng furnace via four subsidiary vacuum chambers, the chambers 
being connected by pairs of rollers through which the strip passes, 
and each subsidiary chamber being pumped by a separate pump. 
After treatment, the strip re-enters the atmosphere via a similar set of 



Furnace chamber 
(•^1,000 'O- 



Strip 



Rollers 




Cooling 
zone Rollers 



Strip 



I \ > > --rt I I 



Supply 
spool 



AA.A/1==...J^AA 



S2 




1,500 mV Roots 



~^ 1^600 m^h"' Roots 



:s5 



150 m^h"' Roots 



Take-up 
spool 



OD 



Gas ballast rotary 



Figure 8.17. Continuous strip vacuum annealing furnace {after Krall, Scheibe and 

Ogiermann'^*') 

chjimbers. The gaps between the rollers and the strips are small 
enough to provide considerable impedance to gas flow, and pressures 
of the order of 10"-^ torr are obtained in the furnace. 



8.6.8 Brazing and Electron Beam Welding 

The advantages of brazing in vacuum are that (f ) fluxes are not 
required since surface oxidation cannot take place, (w) gas inclusions 
are not possible, and {Hi) the brazing alloy wets the base materials and 
flows between them more readily than in atmospheric brazing. 
These factors lead to perfectly clean joints which can be made up to 
50 per cent stronger than those produced by normal methods. The 
elimination of oxygen also means that a far wider range of metals 
may be joined by this technique. 

The brazing alloy should be chosen carefully, and in particular 

I A I.H.V.T. -iOl 



SOME APPLICATIONS OF VACUUM TECHNIQUE 

alloys containing volatile elements {e.g. zinc) should be avoided. A 
range of suitable alloys is given by Belk'^^\ 

The actual brazing technique does not differ greatly from atmos- 
pheric brazing. The structure is assembled in suitable jigs and the 
brazing alloy applied along the joints in wire or strip form. The 
whole assembly is placed in a chamber, whose waUs are water- 
cooled, and which is then evacuated. The assembly is heated to the 
brazing temperature either by radiant heating from electric resistance 
elements mounted within the chamber or by induction heating. 




Figure 8.18. An electron beam welding unit {by courtesy o/'Birvac Ltd.) 

A typical fiimace suitable for vacuum brazing has a working space 
15 in. in diameter and 20 in. high in a chamber 36 in. in diameter and 
60 in. high; it is capable of producing a temperature of 1,400°C by the 
dissipation of 60 kW in molybdenum resistance coUs. It is pumped by 
a 16 in. vapour pimip which gives an effective speed of up to 6,000 
1. sec~^ at working pressures between 10"^ and 10 ~^ torn 

Vacuum welding by electron beam was developed for situations 
in which it was essential that contamination from the heat source and 
surrounding atmosphere was excluded. However, the results of the 
technique are so advantageous that materials which do not neces- 
sarily need such strict conditions are now welded by electron beam. 

202 



THE SIMULATION OF OUTER SPACE AND HIGH ALTITUDE ENVIRONMENTS 

The advantages over normal welding methods include the facts that 
(i) the electron beam is an easily and sensitively controlled source of 
heat, (m) an inert gas is not required for welding reactive metals, (m) 
gas inclusions are avoided, and {iv) the width of the weld can be ex- 
tremely narrow with a penetration depth much greater than by normal 
methods. As an example, 1 in. thick stainless steel has been welded*^®' 
with a depth-to-width ratio of 25 : 1 compared with the normal 
0-5:1. At the other extreme, welding of very thin sheets (less than 
0-01 in.) is possible<^''\ 

A constant velocity beam produced by an electron gun remote 
from the workpiece (see Section 8.6.5) is usually used. The beam is 
focused to a very fine spot on the region to be welded and the work- 
piece is transversed so that the weld line passes through the electron 
beam spot. Weld spot areas as small as 10"'' cm^ in area have been 
obtained, corresponding to a power density of up to 10^ W cm"^. 

The gas load involved in electron beam welding depends largely 
upon the gas content of the workpieces, but in any case is relatively 
small since only a small area of metal is heated at any one time. A 
typical plant"^' has a horizontal cylindrical chamber 2 ft. in diameter 
by 3 ft. in length and has a 20 kV electron gun. Since the process is 
relatively clean, a vapour pump of speed 1,500 1. sec"^ backed by a 
450 1. min~^ rotary pump readily gives working pressures of 10"* torr 
or less. A similar unit is shown in Figure 8.18 where the electron gim 
assembly may be seen projecting vertically from the welding chamber. 



8.7. The Simulation of Outer Space and High 
Altitude Environments 

8.7.1 Outer Space Environment 

This is a particularly interesting application since it can require that 
ultra-high vacuum conditions be produced in chambers of relatively 
large dimensions. 

The simulation of the environment to which a space vehicle may be 
subjected include (i) the pressure believed to obtain in interplanetary 
space (~ 10"" torr), {ii) solar radiation of 0-14W cm"2, {Hi) earth- 
reflected radiation of 0-04 W cm"^, and {iv) a temperature of 3°K. In 
addition,, it is also desirable to simulate the condition that a molecule 
leaving the surface of a space vehicle has only a very small probability 
of returning — outer space acting as an ideal gas sink. 

Details of the simulation of radiation are outside the scope of this 
book and it is sufficient to state that radiation of appropriate spectral 
distribution (usually from discharge lamps) is passed into the chamber 
through vacuum sealed optical windows. 

203 



SOME APPLICATIONS OF VACUUM TECHNIQUE 

The nearest approach to ideal simulation is provided by a chamber 
whose inner walls are entirely lined with liquid helium cryogenic 
pumping surfaces. However, in order to allow the entry of radiation 
it is not possible to obtain complete cryocoverage and it is therefore 
possible for molecules to return to the space vehicle from the radiation 
windows. Santeler^''^' calculated that with 99 per cent cryocoverage 
in a chamber 10 ft. in diameter, only 0-25 per cent of the molecules 
leaving a vehicle in the centre of the chamber return to it, whereas 
with 90 per cent coverage and the same cryopumping speed some 
2-5 per cent of the molecules return. The cryosurfaces must also be 
shielded from the radiation sources in order that the thermal efficiency 
of the former is not impaired, but in such a way that pumping is not 
unduly restricted. 

Steinherz'''^^ describes a chamber whose working volume is 24 in. 
long by 18 in. diameter and which is multiwalled, thus forming a 
series of annular chambers between the main chamber and the 
atmosphere. Initial pump-down is carried out by mechanical and 
vapour pumps giving a pressure of about 10"'' torr after 4 h and the 
innermost walls of the chamber are then baked for about 1 6 h. After 
this the vapour pump traps are filled with liquid nitrogen and liquid 
nitrogen is also circulated through the inner annular chamber, the other 
annular chambers being evacuated, giving a pressure of about 10 "•'^•'^ 
torr after a total of 24 h. The liquid nitrogen flow is then stopped 
and replaced by a circulation of helium gas at 10°K. This results in a 
pressure of about 10"^* torr after a total of 30 h. A major advantage 
of this system of cryopumping is that since the chamber walls are at a 
very low temperature their outgassing is reduced to a very low level. 

For the investigation of many space environment problems ex- 
tremely low pressures and temperatures are not required and much 
valuable work has been carried out at about 10 ~® torr with liquid 
nitrogen-cooled panels for low temperature simulation. Typical of 
such systems is a vessel 25 ft. long by 20 ft. diameter. It is pumped by 
two banks of three parallel connected 32 in. (15,0001. sec"*) vapour 
pumps with liquid nitrogen vapour traps. Each bank is backed by an 
11,0001. min~* rotary pump. Rough pumping to 10~^ torr is 
carried out by two 56,000 1. min"* Roots pumps, each backed by an 
11,0001. min~* rotary pump. A chamber pressure of about 10"^ 
torr is reached after 12 h of pumping. 

8.7.2 High Altitude Environment {low density wind tunnels) 

This requires the production of pressures down to 10 ~^ torr 
(equivalent to ~ 250,000 ft. altitude) with gas velocities to simulate 
flight speeds up to Mach 8. The conditions are achieved by allowing 

204 



references 

gas to enter a previously evacuated chamber, the high speed flow and 
low pressure being maintained by vacuum pumps of sufficient speed. 
Gas entry velocities up to Mach 4 may be obtained by the use of a 
convergent-divergent entry nozzle, and up to Mach 8 by heating the 
gas before its entry to the chamber. The pressure requirements are 
such that gas removal may be affected by vapour booster pumps with 
suitable mechanical force pumps. 

One of the largest of such wind tunnels is pumped by two vapour 
booster pumps in parallel, each capable of a maximum speed of 25,000 
1. sec " * at 5 X 1 " ^ torr. Typical of smaller tunnels is one pim:iped by 
a bank of six pumps giving a potential total speed of 25,000 1. see"*. 



references 

^ Holland, L. Vacuum Deposition of Thin Films. Chapman and Hall, London, 

1961 
^ PiRANi, M. and Yarwood, J. Principles of Vacuum Engineering. Chapman and 

Hall, London, 1961 
^ TuRNBULL, A. H., Barton, R. S. and Riviere, J. C. Introduction to Vacuum 

Technique. Newnes, London, 1962 
* Langmuir, I. Phys. Rev. 2 (1913) 329 
^ PiRANi, M. and Yarwood, J. Principles of Vacuum Engineering, pp 314-16. 

Chapman and Hall, London, 1961 
» Olsen, L. O., Smith, C. S. and Critenden, E. C. J. appl. Phys. 16 (1945) 425 
■^ SiDDALL, G. and Probyn, B. A. Vacuum Symp. Trans. (1962) 1017 
8 Clegg, P. L. and Crook, A. W. J. scient. Instrum. 29 (1952) 201 
' Turner, J. A., Birtwistle, J. K. and Hoffmann, G. R. J. scient. Instrum. 

40 (1963) 557 
"> Frauenfelder, H. Helv. phys. Acta 27 (1950) 347 
" Scott, G. O., McLauchlan, T. A. and Sennett, R. G. J. appl. Phys. 21 

(1950) 843 
" Steckelmacher, W. and English, J. Vacuum Symp. Trans. (1962) 852 
" Hass, G. and Scott, N. W. J. opt. Soc. Am. 39 (1949) 179 
" Holland, L. and Hacking, K. Electron. Engng 26 (1954) 296 
" Holland, L. Vacuum 6 (1959) 161 
18 Barker, D. W. Int. Plast. Engng 3 (1963) 1 16 
I'' Williams, B. J. Electroplg Metal Finish. 13 (1960) 247 

18 Ward, L. Appl. Mater. Res. 3 (1964) 163 

19 SiDDALL, G. and Probyn, B. A. Radio electron. Compon. 2 (1961) 277 
=0 SiDDALL, G. and Probyn, B. A. Br. J. appl. Phys. 12 (1961) 668 

" SiDDALL, G. Vacuum 9 (1959) 274 

22 DuTHiE, A., Humphreys, S. and Probyn, B. A. Electron. Engng 35 (1963) 430 

28 Kay, D., Techniques for Electron Microscopy, p 271. Blackwell, Oxford, 1961 

" Robinson, N. W. Research 15 (1962) 413 

28 Belk, J. A. Vacuum Techniques in Metallurgy. Pergamon and MacmiUan, 

London and New York, 1963 
28 Kino, J. B. Vacuum Metallurgy, pp 35-58. Rheinhold, New York, 1958 
^■' Iron Age, 182 (1958) 81 

28 Matejka, W.J. Vacuum Symp. Trans. (1962) 757 

29 Wright, W. J. and Harper, H. E. Nucl. Engng, Lend. 6 (1961) 344 
88 Private communication, I MI Ltd. 

81 Smith, H. R., Hunt, C. d'A. and Hanks, C. W. Vacuum Symp. Trans. (1962) 
708 

205 



SOME APPLICATIONS OF VACUUM TECHNIQUE 

" St Clair, H. W. Vacuum Metallurgy, pp 295-305. Rheinhold, New York, 

1958 
" Metal Prog. 76 (1959) 111 

'* Krall, F., Scheibe, W. and Ogiermann, W. Vacuum Symp. Trans. (1962) 791 
'• Belk, J. A. Vacuum Techniques in Metallurgy, p 209. Pergamon and Macmillan, 

London and New York, 1963 
''« ScHWARZ, H. Vacuum Symp. Trans. (1962) 699 
" Harper, M. E. and Nunn, E. G. Br. Weld. J. 7 (1960) 331 
'* Santeler, D. J. 'Theory and Design of Cryogenic Pumping Systems for 

Space Environmental Simulators'. 2nd Ann. Symp. Space Vacuum Simulation, 

Atmospheric Devices Laboratory, 1961 
*" Steinherz, H. a. Handbook of High Vacuum Engineering, Chap 8. Rheinhold, 

New York, 1963 



206 



LIST OF PRINCIPAL SYMBOLS 



A 
a 
B 
b 
C 

c„, 
c, 

c 

Ca 

D 

D, 
E 
e 
F 

f 

/o 

H 

Ha 

h 

J 
K 

Kn 

k 
L 
Lo 

I 

M 

m 

N 
K 
No 

n 
ria 



Area 

Condensation coefficient 

Magnetic field strength 

Solubility (of a gas in a solid) 

Conductance (in gaseous flow) 

Conductance in molecular flow 

Conductance in transitional flow 

Specific heat of a gas at constant volume 

Conductance in viscous flow 

Root mean square velocity of gas molecules 

Arithmetic mean velocity of gas molecules 

Diameter (of pipes, orifices, etc.) 

Diffusion coefficient (of gases in a solid) 

Electric field strength 

Charge on an electron (1-60 x 10"^^ coulomb) 

Deflection of gauge indicator 

Coverage (fraction of possible adsorption sites actually 

occupied by molecules) 
Value ofy"at time zero 

Numerical factors (in conductance calculations) 
Activation energy of absorption (of a gas in a solid) 
Heat of desorption (of adsorbed gas from a solid) 
Number of hours 
Dissociation constant (of a gas) 
Outgassing constant 
Value of K after h hours 
Ultimate value of K 

Boltzmann's constant (1-38 x 10"" erg "K"!) 
Length (of pipe, etc.) 
Thickness (of vessel wall) 
Length (of gas or liquid column) 
Molecular weight 
Mass of a molecule 
Mass of an ion 

Total number of gas molecules in a given volume 
Avogadro's number (6-06 x 10^^) 
Number of gas molecules in a monolayer or unit area of 

surface 
Number of molecules per unit volume 
Number of molecules adsorbed on unit area of surface 



207 



LIST OF PRINCIPAL SYMBOLS 



P 

P 
P 

Ap 
Pi, 

Pm 
Pu 
Pv 
pu, 

9 

R 

Re 

Rpt 
Rv 

r, 
S 

So 

s. 

s„ 

So 

s 

T 

t 

t. 

to 
U 
V 

V 

W 

z 

z 

Y 

e 

V 

V» 



e 

K 

A 



R. 



2.3,4 



Number of molecules desorbed from unit area of surface 

Permeation constant 

Gas pressure 

Mean pressure 

Minimum detectable pressure change 

Critical backing pressure 

Pressure at which speed of a vapour pump becomes a maximum 

Ultimate pressure 

Partial pressure of vapour 

Working pressure 

Throughput (of gas) 

Throughput due to outgassing 

True leak rate 

Gas constant per mole (8-316 x W erg mole"^ °K"^) 

Rate of gaseous diffusion 

Rate of evaporation 

Resistances 

Maximum mass flow rate of vapour 

Radius 

Speed (of pumping) 

Maximum speed of a cryopump 

Effective speed (accounting for effect of impedance to flow) 

Maximum speed of a vapour pump 

Volumetric displacement (of mechanical pumps) 

Sticking coefficient 

Absolute temperature (°K) 

Time 

Sojourn time (average time spent by a given molecule on a 

particular adsorption site) 
Time constant in outgassing equation 
Perimeter 
Volume 

Velocity, general 
Potential difference 
Impedance (to gaseous flow) 
Exponent in vapour pump speed equation 
Numerical factors (in conductance calculations) 
Rate determining exponent in outgassing equations 
Gas concentration (in a solid) 
Viscosity (of gases) 
Viscosity at temperature 6 

Geometrical constant (in leak detection equations) 
Tenaperature (°C) 
Thermal conductivity (of a gas) 
Mean free path of gas molecules 
Number of molecular bombardments on unit area of surface 

in unit time 



208 



LIST OF PRINCIPAL SYMBOLS 

p Density (of liquids or gases) 

a Diameter of a molecule 

<f> Substitution sensitivity factor (in leak detection equations) 

ip Gauge sensitivity factor 

0)0 Angular velocity (of ions) 



209 



INDEX 



Absorption, 10-12, 17 

energy of, 13 
Activation, 

heat of, 10 

of cathode, 184, 186, 187 
Admittance area, of vapour pump, 53, 

54 
Adsorption, 10-12, 17 

heat of, 10 

isobar, 12 

isotherm, 12, 19 

physical, 10, 19, 68 

rate of, 1 1 
Air admittance valve — see Valve 
Aluminium, evaporation of, 170, 176, 

177 
Annealing, in vacuum, 200, 201 
Applications of vacuum technology, 

170-205 
Arc melting, 194-197 

Back 

diffusion, 53, 56, 59 

streaming, 55 

reduction of, 123, 128 
Backing, 52, 56 

space, 162 
Baffle, 

chevron, 129 

cold, 55, 129 

glass, 131 

refrigerated, 8, 57, 157 

valve, 123, 155, 156 
Baking, to reduce outgassing, 15, 20, 

139, 154, 156, 157, 187 
Bayard-Alpert gauge, 95 
Bellows, metal, 119 
Blooming of lenses, 1 70, 179 
Bombardment, 

ionic, 22, 186 

molecular, 4 
Bourdon gauge, 72, 73 
Boyle's Law, 2, 8 
Brazing, in vacuum, 201, 202 
Butterfly valve, 123, 124 



Calibration of gauges — see Gauges 
Capacitors, manufacture of, 178, 179 
Capacity, sorptive, 60 
Carbon, deposition of, 181 
Cathode activation, 184, 186, 187 
Cathode ray tubes, manufacture of, 186 
Charcoal, in sorption pumps, 61 
Chemisorption, 10, 12, 19 

activated, 10, 11, 15 
Circuit components, production of, 1 79, 

180 
Coating, 

corrosion resistance, 178 
decorative, 178 
in vacuum, 170-181 
Cold trap, 
design of, 128 

effect on gauge readings, 109 
glass, 131 
uses of, 46, 55, 57 
Computer elements, manufacture of, 

180 
Concentration, of gas in solid, 12, 13, 17 
Condensation, coefficient, 62 
Conductance, 24 

effect on pumping speed, 25, 140 
in parallel, 25 
in scries, 25 
of baffles, 36 
Conductance, 

molecular, 28, 29, 33 
annular orifice, 34 
bend, 35, 36 
concentric cylinders, 34 
effects of gas, 29 
effect of temperature, 29 
Monte Carlo calculation, 35, 36 
orifice, 30, 31 
rectangular duct, 34 
total, 28, 31, 34, 35 
viscous, 27-30 
effect of gas, 30 
effect of temperature, 30 
Conductivity, thermal, 7, 86, 164 
Construction of system — see Systems 



211 



INDEX 



Couplings, demountable, 115, 116 

Coverage, 11, 19 

Creep, oil, 128 

Critical backing pressure — see Pump 

and Pressure 
temperature, 8 
Crucibles, as evaporant sources, 174, 

175 
Cryopump — see Pump 
Cryotrapping, 63 
Cut-off, mercury, 131 

Decorative coating, 178 
Degassing, 

by baking, 114, 132 

circulation, 190 

progressive, 189 

static, 189 

stream, 191 

vacuum, 188-191 
Density, molecular, 3, 5 
Design, 

ofvacuum systems, 134-154 

numerical, 139-146 
examples of, 146-154 
Desorption, 10, 12, 13, 159 — see also 
Outgassing 

energy of, 10, 11,22 

heat of, 10 

rate of, 1 1 
Diameter, molecular, 5 
Diffusion, 

of gas, 13 

back, 53, 56, 59 
Discharge, lamps, manufacture of 183, 
184 

tube — see Gauges and Leak detection 
Dissociation, 13 

products, 105 
Distillation, metal purification by, 200 

Electron 
beam, 

heating, 175 
melting, 197-201 
welding, 202, 203 
microscope, specimens for, 180 
tubes, manufacture of, 184-188 
Electrode, 132, 133 
Evaporant sources, 172, 173, 179 
Evaporation, 

by flash filament, 1 75 
by sputtering, 1 75, 1 76 



Evaporation — contd. 

data for various materiab, 174 

equipment for, 171-175 

of various materiab, 170-181 

rate of, 171 
Evapor-Ion pump — see Pump 

Fernico, glass/metal seal, 132 
Film, 

epitaxial, 180 

metallic oxide, 176 

optical, 179 

thickness measurement of, 176 

thin, 180 
Filter, 

interference, production of, 179 

heat, production of, 179 
Flow, gaseous, 

laminar, 6, 27 

mechanisms of, 26 

molecular, 7, 26, 27, 141 

rate of, 2 

theory of, 24-36 

transitional, 28, 29, 183, 141 

turbulent, 26, 27 

viscous, 26, 27, 29, 143, 159, 183 
Fluid, vapour pump, 54, 55 
Fluorescence, 79 
Fractionating pump, 55, 56 
Freon, use in leak detection, 161, 167 
Furnace, 

arc, 194-197 

electron beam, 197-200 

induction, 192-194 

Gas, 

ballast, 46-49, 138 
theory of, 48^9 
valve, 38, 47, 48 
laws, 1 

load, 139, 141 
probe, 160, 161, 163 

detection of, 164-167, 168 
Gaseous flow — see Flow 
Gauge, 

Alphatron, 98 

aneroid, 73, 74 

Bayard-AIpert, 95 

Bourdon, 72, 73 

calibration of, 76, 85, 89, 93, 94, 97, 

106, 108 
discharge tube, 74, 75 
leak detector, 164 



212 



INDEX 



Gauge — contd. 

interpretation of readings, 108, 109 
ionization, 91-101 

cold cathode, 95-98, 99, 100 

hot cathode, 91-94 
calibration, 93, 94 
control circuit, 93 
pressure limits, 94 
x-ray limit, 94 

leak detector, 163, 165 

modulated, 101 
Knudsen, 85 
Lafferty, 98 
magnetron, 98, 99 
manometer, 71-75, 108, 183 

differential, 72 
mass spectrometer, 101-106, 109 

leak detector, 167 

magnetic deflection, 101, 102 

monopole, 104 

Omegatron, 103 

quadrupole, 104 

trochoidal, 103 
McLeod, 75-85 

bench model, 81 

calibration, 76, 85 

effect of outgassing, 79 

effect of vapour streaming, 80 

effect of vapours, 78 

Measuvac, 82 

mercury reservoir, 80 

multi-range, 83 

Vacustat, 82 
Penning, 95-97 

calibration, 97 

control circuit, 97 

low pressure starting, 97 

pressure limits, 97 
Pirani, 86-90 

calibration, 89 

control circuit, 88, 89 

leak detector, 164 

multi-head, 90 

pressure limits, 88 
radiometer, 85 
Redhead, 98, 99 
semi-conductor, 90 
sensitivity (in leak detection), 162, 

165 
thermal conductivity, 86-9 1 
thermocouple, 90 
Gettering, 21, 22, 66, 77, 183, 184, 186, 

187 



Halide leak detector, 161, 167 
Heat treatment of metals, 200, 201 
High altitude, simulation of, 204, 205 
Hydrogen, 

desorption of, 109 

outgassing, 14, 18 

Impedance, 24 

Induction melting, 192-194 

Ion 

current, 91, 93, 94, 96 

in magnetic and electric fields, 9 
pump — see Pump 
Ionization, 8, 9, 75, 91 

energy, 9, 93, 165 

gauge — see Gauge 

multiple, 105 

secondary, 10 
Isotopes, 9, 106 

Joints, 
cone, 130 

demountable, 114, 115, 132 
high temperature, 116-118 
permanent, 112-114 
soldered, 1 12 

ultra high vacuum, 116-118 
welded, 112-114 

Kinetic theory, 3, 4, 6, 7 
Kinney pump— see Pumps 
Knudsen gauge, 85 
Kovar, glass/metal seal, 132, 133 

Lafferty gauge, 98 
Leak, 

detection, 160-168, 183, 188 

by discharge tube, 164, 167 

by halide, 161, 167 

by immersion, 160 

by ion gauge, 165, 167 

by Palladium barrier, 166, 167 

by Pirani gauge, 164, 167 

by mass spectrometer, 167 

general procedure, 168 

system, 162 
rate, definition, 158 

minimum detectable, 164 
real, 136, 158, 159 
true, 136, 158-160 
virtual, 136, 158, 159 



213 



INDEX 



Manometer — see Gauge 
Mass spectra, 104, 105 

spectrometer — see Gauge 
Materials of system construction, 111, 

112 
McLeod gauge — see Gauge 
Mean free path, 4, 5, 27, 75 
Measuvac gauge — see Gauge 
Metallurgy, vacuum, 188-203 
Metals, deposition of, 170 
Melting in vacuum, 192-200 

by electric arc, 194-197 

by electron beam, 197-200 

by induction heating, 192-194 
Micro-circuits, production of, 179, 180 
Mirrors, aluminizing of, 170, 176, 177 
Molecular, 

density, 3, 5 

diameter, 5 

flow — see Flow 

sieve, 61, 63 
Monolayer, time to form, 1 1 
Monte Carlo calculation of conduc- 
tance, 35 
Molybdenum boat, use in evaporation, 
173-175, 179 

Needle valve, 126 
Neoprene, 114 

outgassing of, 16 
Non-metals, evaporation of, 170, 174 
177, 179 

Oil 

creep, 128 
rotary pump, 41 
vapour pump, 55 
Omegatron, 103 

Orifice, conductance of — see Conduc- 
tance 
O-ring seals — see Seal 
Outer space, simulation of, 203, 204 
Outgassing, 13-21, 136, 137, 159 
constant, 13 

values of, 16 
during pump dovm, 144 
effect of baking, 15,20, 139, 154, 156, 
157 
of cleaning and polishing, 1 5 
on McLeod gauge, 79 
on ultimate pressure, 57, 146, 152 
on ultra high vacuum, 1 46, 1 52 
of pipelines, 141 



Outgassing — contd. 
theory, 16-21 

adsorption controlled, 19 
diffusion controlled, 17, 18 
effect of temperature, 20, 21 
permeation, 18 
water vapour, 19 
variation with pumping time, 14, 15 
Oxide, films of, 176 

Palladium barrier, 166 
Paper, metallizing of, 178 
Penning gauge — see Gauge 
Permeation, 12, 13 

constant, 17 
Phosphorus pentoxide, 46 
Pipeline, 

conductance of — see Conductance 

dimensions, estimation of, 140 

outgassing of, 141 
Pirani gauge — see Gauge 
Pressure, 3 

absolute, 1 

critical backing, 56, 59, 60, 142 

fluctuations, 164 

measurement, 71-109 

partial, 5, 109 

saturation vapour, 8, 45, 79 

ultimate, 140, 158 
of pumps — see Pump 

units, 1 

vapour, 8 

working, 136, 139 
Probe gas — sec Gas 
Pump, 37-69 

booster — see Pump, vapour booster 

combinations, 138, 139 

cryogenic, 8, 62-65, 139 
characteristics, 63 
speed, 62-64 
ultimate pressure, 62 

displacement, 41 

down time, 144, 145 

evapor-ion, 66, 67, 139 

fore, 134, 142 

getter, 65-69, 139 

ion, 65-69, 139 

Kinney, 43-44 

main, 134, 137, 140 

mechanical, 37-52 
characteristics, 49, 50 
displacement, 41 
effect of vapour, 45, 46 



214 



INDEX 



Pump — contd. 
mechanical — contd. 
speed, 45, 50 
two stage, 45 

ultimate pressure, 44, 45, 52 
oil sealed rotary, 37-50, 138, 139 
Penning, 65 
Roots, 50, 51, 138 
rotating vane type, 37-42 
sliding vane type, 42, 43 
sorption, 60-62, 138, 139 

speed, 60, 61 
sputter-ion, 67-69 
vapour, 52-60, 138, 139 
action of, 52-54 
backing of, 56 
booster, 58-60, 138 
characteristics, 59, 60 
critical backing pessure, 60 
speed, 59, 60 
ultimate pressure, 59, 60 
characteristics, 56, 57 
cooling of, 56 
fluid, 54, 55 
fractionating, 55, 56 
multi-stage, 53, 54 
roughing of, 56 
single stage, 52 
speed, 53 

maximum, 56 
ultimate pressure, 56, 57 
Pumping assembly, 135 



Seal — contd. 

permanent, 112-114 

rotating, 118-120 

sliding, 120 

ultra-high vacuum, 116-118 

Wilson, 119, 123 

wobble, 120 
Self-fractionation, 55 
Semi-conductor gauge, 90 
Silica gel, 46 

Solubility, of gas in solid, 12, 13 
Sorption, 10, 20 

pump — see Pump 
Spectra, mass, 103-105 
Spectrometer, mass — see Gauge 
Speed, 2 

effective, 25, 26, 140, 141, 158 

of pumps — see Pump 
Sputter-ion pumps — see Pump 
Sputtering, 22, 170, 175, 187 
Sticking coeflScient, 1 1 

probability, 11, 36 
Stopcocks, glass, 130 
Streaming, back, 55 
Substitution sensitivity factor, 163 
Substrate, cleaning of, 1 72 
System, 

construction of, 111-133 

design of, 134-154 

glass. 111, 129-132 

glass/metal, 131 

operation of, 154-157 



Radiation, thermal, 87 

Radio valves, manufacture of, 184-186 

Redhead gauge, 98, 99 

Roughing, 56 

Run-away, in sputter pump, 68 

Seal, 

bakeable, 116-118 

bellows, 119 

ceramic/metal, 132 

copper washer, 118 

demountable, 114-118, 120, 132 

gasket, 114-118 

glass, 120 

demountable, 120 
glass/metal, 132 
graded, 130 

high temperature, 116-118 
O-ring, 114-116 



Temperature, critical, 8 
Thermal conductivity, 7, 86, 164 
gauge — see Gauge 

radiation, 87 
Thermocouple gauge — see Gauge 
Thickness of film, measurement, 176 
Throughput, 2 

due to outgassing, 136 

due to process, 137 
Time, pump down, 144, 145 

sojourn, 11 
Torr, definition, 1 

Travelling wave tube, manufacture, 187 
Tungsten filament lamp, 

manufacture, 181-183 

wire, in evaporation, 174, 176 



Ultimate pressure — see Pressure 



215 



INDEX 



acustat gauge — see Gauge 
Vacuum 

coating, 170-181 

degassing of metals, 188-191 

progressive, 189, 190 

purification by, 191 

static, 189 

stream, 191 
Valve, 

air admittance, 42, 125, 135, 146 
baffle, 123, 135, 155, 156 
bellows, 122 
butterfly, 123, 124 
conductance of, 

closed — see appropriate valve 

open — see appropriate valve 
diaphragm, 150 
fast closing, 108 
gate, 124, 127 
glass, 130 

high vacuum, 121-124, 127 
isolation, 147, 151 
needle, 126 

magnetically operated, 125 
medium vacuum, 121 
pneumatically operated, 125 
pressure sensitive, 127 
throttle, 162 

ultra-high vacuum, 126, 132 
Vapour, 

condensation in oil sealed pump, 45 
pressure, saturated, 8, 45, 79 



Vapour — -contd. 

removal from oil sealed pump, 46-49 

pump — see Pump 
Vapours, 7 
Velocity, 

arithmetic, 4 

root mean squares, 3 
Viscosity, gaseous, 6 

values 30 
Viscous 

flow — see Flow 

conductance — see Conductance 
Viton, 114 

outgassing of, 16 
Voltage stabilizer tube, 186 

Weld, double, 113 
Welding, 

argon arc, 1 12, 1 14 

electron beam, 201-203 
Wilson seal, 119, 123 
Wobble seal, 120 
Working, 

pressure, 136, 139 

vessel, 134, 136, 172 

X-ray 

limit, ion gauges, 94, 95, 98, 101 
tube manufacture of, 186 



Zeolite, 61 



216